[v2,4/8] ipa: libipa: Add MeanLuminanceAgc base class

19896 diff mbox series

The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
very large part; following the Rpi IPA's algorithm in spirit with a
few tunable values in that IPA being hardcoded in the libipa ones.
Add a new base class for MeanLuminanceAgc which implements the same
algorithm and additionally parses yaml tuning files to inform an IPA
module's Agc algorithm about valid constraint and exposure modes and
their associated bounds.

Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
---
Changes in v2:

	- Renamed the class and files
	- Expanded the documentation
	- Added parseTuningData() so derived classes can call a single function
	  to cover all the parsing in ::init().

 src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
 src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
 src/ipa/libipa/meson.build            |   2 +
 3 files changed, 674 insertions(+)
 create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
 create mode 100644 src/ipa/libipa/agc_mean_luminance.h

Hi Dan,

thank you for the patch.

On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> very large part; following the Rpi IPA's algorithm in spirit with a
> few tunable values in that IPA being hardcoded in the libipa ones.
> Add a new base class for MeanLuminanceAgc which implements the same
> algorithm and additionally parses yaml tuning files to inform an IPA
> module's Agc algorithm about valid constraint and exposure modes and
> their associated bounds.
> 
> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> ---
> Changes in v2:
> 
> 	- Renamed the class and files
> 	- Expanded the documentation
> 	- Added parseTuningData() so derived classes can call a single function
> 	  to cover all the parsing in ::init().
> 
>  src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
>  src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
>  src/ipa/libipa/meson.build            |   2 +
>  3 files changed, 674 insertions(+)
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.h
> 
> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> new file mode 100644
> index 00000000..02e223cf
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> @@ -0,0 +1,581 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> + */
> +
> +#include "agc_mean_luminance.h"
> +
> +#include <cmath>
> +
> +#include <libcamera/base/log.h>
> +#include <libcamera/control_ids.h>
> +
> +#include "exposure_mode_helper.h"
> +
> +using namespace libcamera::controls;
> +
> +/**
> + * \file agc_mean_luminance.h
> + * \brief Base class implementing mean luminance AEGC
> + */
> +
> +namespace libcamera {
> +
> +using namespace std::literals::chrono_literals;
> +
> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> +
> +namespace ipa {
> +
> +/*
> + * Number of frames for which to run the algorithm at full speed, before slowing
> + * down to prevent large and jarring changes in exposure from frame to frame.
> + */
> +static constexpr uint32_t kNumStartupFrames = 10;
> +
> +/*
> + * Default relative luminance target
> + *
> + * This value should be chosen so that when the camera points at a grey target,
> + * the resulting image brightness looks "right". Custom values can be passed
> + * as the relativeLuminanceTarget value in sensor tuning files.
> + */
> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> +
> +/**
> + * \struct AgcMeanLuminance::AgcConstraint
> + * \brief The boundaries and target for an AeConstraintMode constraint
> + *
> + * This structure describes an AeConstraintMode constraint for the purposes of
> + * this algorithm. The algorithm will apply the constraints by calculating the
> + * Histogram's inter-quantile mean between the given quantiles and ensure that
> + * the resulting value is the right side of the given target (as defined by the
> + * boundary and luminance target).
> + */
> +
> +/**
> + * \enum AgcMeanLuminance::AgcConstraint::Bound
> + * \brief Specify whether the constraint defines a lower or upper bound
> + * \var AgcMeanLuminance::AgcConstraint::lower
> + * \brief The constraint defines a lower bound
> + * \var AgcMeanLuminance::AgcConstraint::upper
> + * \brief The constraint defines an upper bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::bound
> + * \brief The type of constraint bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qLo
> + * \brief The lower quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qHi
> + * \brief The upper quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::yTarget
> + * \brief The luminance target for the constraint
> + */
> +
> +/**
> + * \class AgcMeanLuminance
> + * \brief A mean-based auto-exposure algorithm
> + *
> + * This algorithm calculates a shutter time, analogue and digital gain such that
> + * the normalised mean luminance value of an image is driven towards a target,
> + * which itself is discovered from tuning data. The algorithm is a two-stage
> + * process.
> + *
> + * In the first stage, an initial gain value is derived by iteratively comparing
> + * the gain-adjusted mean luminance across an entire image against a target, and
> + * selecting a value which pushes it as closely as possible towards the target.
> + *
> + * In the second stage we calculate the gain required to drive the average of a
> + * section of a histogram to a target value, where the target and the boundaries
> + * of the section of the histogram used in the calculation are taken from the
> + * values defined for the currently configured AeConstraintMode within the
> + * tuning data. This class provides a helper function to parse those tuning data
> + * to discover the constraints, and so requires a specific format for those
> + * data which is described in \ref parseTuningData(). The gain from the first
> + * stage is then clamped to the gain from this stage.
> + *
> + * The final gain is used to adjust the effective exposure value of the image,
> + * and that new exposure value is divided into shutter time, analogue gain and
> + * digital gain according to the selected AeExposureMode. This class expects to
> + * use the \ref ExposureModeHelper class to assist in that division, and expects
> + * the data needed to initialise that class to be present in tuning data in a
> + * format described in \ref parseTuningData().
> + *
> + * In order to be able to derive an AGC implementation from this class, an IPA
> + * needs to be able to do the following:
> + *
> + * 1. Provide a luminance estimation across an entire image.
> + * 2. Provide a luminance Histogram for the image to use in calculating
> + *    constraint compliance. The precision of the Histogram that is available
> + *    will determine the supportable precision of the constraints.
> + */
> +
> +AgcMeanLuminance::AgcMeanLuminance()
> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> +{
> +}
> +
> +/**
> + * \brief Parse the relative luminance target from the tuning data
> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> + */
> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> +{
> +	relativeLuminanceTarget_ =
> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> +}
> +
> +/**
> + * \brief Parse an AeConstraintMode constraint from tuning data
> + * \param[in] modeDict the YamlObject holding the constraint data
> + * \param[in] id The constraint ID from AeConstraintModeEnum
> + */
> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> +{
> +	for (const auto &[boundName, content] : modeDict.asDict()) {
> +		if (boundName != "upper" && boundName != "lower") {
> +			LOG(AgcMeanLuminance, Warning)
> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> +			continue;
> +		}
> +
> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> +		double qLo = content["qLo"].get<double>().value_or(0.98);
> +		double qHi = content["qHi"].get<double>().value_or(1.0);
> +		double yTarget =
> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> +
> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> +
> +		if (!constraintModes_.count(id))
> +			constraintModes_[id] = {};
> +
> +		if (idx)
> +			constraintModes_[id].push_back(constraint);
> +		else
> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> +	}
> +}
> +
> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableConstraintModes;
> +
> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> +	if (yamlConstraintModes.isDictionary()) {
> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> +			if (AeConstraintModeNameValueMap.find(modeName) ==
> +			    AeConstraintModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown constraint mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeDict.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid constraint mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			parseConstraint(modeDict,
> +					AeConstraintModeNameValueMap.at(modeName));
> +			availableConstraintModes.push_back(
> +				AeConstraintModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If the tuning data file contains no constraints then we use the
> +	 * default constraint that the various Agc algorithms were adhering to
> +	 * anyway before centralisation.
> +	 */
> +	if (constraintModes_.empty()) {
> +		AgcConstraint constraint = {
> +			AgcConstraint::Bound::lower,
> +			0.98,
> +			1.0,
> +			0.5
> +		};
> +
> +		constraintModes_[controls::ConstraintNormal].insert(
> +			constraintModes_[controls::ConstraintNormal].begin(),
> +			constraint);
> +		availableConstraintModes.push_back(
> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> +	}
> +
> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> +
> +	return 0;
> +}
> +
> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableExposureModes;
> +
> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> +	if (yamlExposureModes.isDictionary()) {
> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> +			if (AeExposureModeNameValueMap.find(modeName) ==
> +			    AeExposureModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown exposure mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeValues.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid exposure mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<uint32_t> shutters =
> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> +			std::vector<double> gains =
> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> +
> +			if (shutters.size() != gains.size()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain array sizes unequal";
> +				return -EINVAL;
> +			}
> +
> +			if (shutters.empty()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain arrays are empty";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<std::pair<utils::Duration, double>> stages;
> +			for (unsigned int i = 0; i < shutters.size(); i++) {
> +				stages.push_back({
> +					std::chrono::microseconds(shutters[i]),
> +					gains[i]
> +				});
> +			}
> +
> +			std::shared_ptr<ExposureModeHelper> helper =
> +				std::make_shared<ExposureModeHelper>();
> +			helper->init(stages);
> +
> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If we don't have any exposure modes in the tuning data we create an
> +	 * ExposureModeHelper using an empty vector of stages. This will result
> +	 * in the ExposureModeHelper simply driving the shutter as high as
> +	 * possible before touching gain.
> +	 */
> +	if (availableExposureModes.empty()) {
> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> +		std::vector<std::pair<utils::Duration, double>> stages = { };
> +
> +		std::shared_ptr<ExposureModeHelper> helper =
> +			std::make_shared<ExposureModeHelper>();
> +		helper->init(stages);
> +
> +		exposureModeHelpers_[exposureModeId] = helper;
> +		availableExposureModes.push_back(exposureModeId);
> +	}
> +
> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> + * \param[in] tuningData the YamlObject representing the tuning data
> + *
> + * This function parses tuning data to build the list of allowed values for the
> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> + * the data in a specific format; the Agc algorithm's tuning data should contain
> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> + * Each mode dict may contain either a "lower" or "upper" key or both, for
> + * example:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeConstraintMode:
> + *         ConstraintNormal:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *         ConstraintHighlight:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *           upper:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.8
> + *
> + * \endcode
> + *
> + * For the AeExposureMode control the data should contain a dictionary called
> + * AeExposureMode containing per-mode setting dictionaries with the key being a
> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> + * contain an array of shutter times with the key "shutter" and an array of gain
> + * values with the key "gain", in this format:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeExposureMode:
> + *         ExposureNormal:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *         ExposureShort:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *
> + * \endcode
> + *
> + * @return 0 on success or a negative error code
> + */
> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> +{
> +	int ret;
> +
> +	parseRelativeLuminanceTarget(tuningData);
> +
> +	ret = parseConstraintModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	ret = parseExposureModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief configure the ExposureModeHelpers for this class
> + * \param[in] minShutter Minimum shutter time to allow
> + * \param[in] maxShutter Maximum shutter time to allow
> + * \param[in] minGain Minimum gain to allow
> + * \param[in] maxGain Maximum gain to allow
> + *
> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> + * ExposureModeHelper that has been created for this class.
> + */
> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> +						    utils::Duration maxShutter,
> +						    double minGain,
> +						    double maxGain)
> +{
> +	for (auto &[id, helper] : exposureModeHelpers_)
> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::constraintModes()
> + * \brief Get the constraint modes that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::exposureModeHelpers()
> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::controls()
> + * \brief Get the controls that have been generated after parsing tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> + * \brief Estimate the luminance of an image, adjusted by a given gain
> + * \param[in] gain The gain with which to adjust the luminance estimate
> + *
> + * This function estimates the average relative luminance of the frame that
> + * would be output by the sensor if an additional \a gain was applied. It is a
> + * pure virtual function because estimation of luminance is a hardware-specific
> + * operation, which depends wholly on the format of the stats that are delivered
> + * to libcamera from the ISP. Derived classes must implement an overriding
> + * function that calculates the normalised mean luminance value across the
> + * entire image.
> + *
> + * \return The normalised relative luminance of the image
> + */
> +
> +/**
> + * \brief Estimate the initial gain needed to achieve a relative luminance
> + * target
> + *
> + * To account for non-linearity caused by saturation, the value needs to be
> + * estimated in an iterative process, as multiplying by a gain will not increase
> + * the relative luminance by the same factor if some image regions are saturated
> + *
> + * \return The calculated initial gain
> + */
> +double AgcMeanLuminance::estimateInitialGain()
> +{
> +	double yTarget = relativeLuminanceTarget_;
> +	double yGain = 1.0;
> +
> +	for (unsigned int i = 0; i < 8; i++) {
> +		double yValue = estimateLuminance(yGain);
> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> +
> +		yGain *= extra_gain;
> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> +				<< ", Y target: " << yTarget
> +				<< ", gives gain " << yGain;
> +
> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> +			break;
> +	}
> +
> +	return yGain;
> +}
> +
> +/**
> + * \brief Clamp gain within the bounds of a defined constraint
> + * \param[in] constraintModeIndex The index of the constraint to adhere to
> + * \param[in] hist A histogram over which to calculate inter-quantile means
> + * \param[in] gain The gain to clamp
> + *
> + * \return The gain clamped within the constraint bounds
> + */
> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> +					     const Histogram &hist,
> +					     double gain)
> +{
> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> +	for (const AgcConstraint &constraint : constraints) {
> +		double newGain = constraint.yTarget * hist.bins() /
> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> +
> +		if (constraint.bound == AgcConstraint::Bound::lower &&
> +		    newGain > gain)
> +			gain = newGain;
> +
> +		if (constraint.bound == AgcConstraint::Bound::upper &&
> +		    newGain < gain)
> +			gain = newGain;
> +	}
> +
> +	return gain;
> +}
> +
> +/**
> + * \brief Apply a filter on the exposure value to limit the speed of changes
> + * \param[in] exposureValue The target exposure from the AGC algorithm
> + *
> + * The speed of the filter is adaptive, and will produce the target quicker
> + * during startup, or when the target exposure is within 20% of the most recent
> + * filter output.
> + *
> + * \return The filtered exposure
> + */
> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> +{
> +	double speed = 0.2;
> +
> +	/* Adapt instantly if we are in startup phase. */
> +	if (frameCount_ < kNumStartupFrames)
> +		speed = 1.0;
> +
> +	/*
> +	 * If we are close to the desired result, go faster to avoid making
> +	 * multiple micro-adjustments.
> +	 * \todo Make this customisable?
> +	 */
> +	if (filteredExposure_ < 1.2 * exposureValue &&
> +	    filteredExposure_ > 0.8 * exposureValue)
> +		speed = sqrt(speed);
> +
> +	filteredExposure_ = speed * exposureValue +
> +			    filteredExposure_ * (1.0 - speed);
> +
> +	return filteredExposure_;
> +}
> +
> +/**
> + * \brief Calculate the new exposure value
> + * \param[in] constraintModeIndex The index of the current constraint mode
> + * \param[in] exposureModeIndex The index of the current exposure mode
> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> + *	      the calculated gain

nit: no indentation

> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> + *	      statistics in use derive

nit: no indentation

> + *
> + * Calculate a new exposure value to try to obtain the target. The calculated
> + * exposure value is filtered to prevent rapid changes from frame to frame, and
> + * divided into shutter time, analogue and digital gain.
> + *
> + * \return Tuple of shutter time, analogue gain, and digital gain
> + */
> +std::tuple<utils::Duration, double, double>
> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> +				 uint32_t exposureModeIndex,
> +				 const Histogram &yHist,
> +				 utils::Duration effectiveExposureValue)
> +{
> +	/*
> +	 * The pipeline handler should validate that we have received an allowed
> +	 * value for AeExposureMode.
> +	 */
> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> +		exposureModeHelpers_.at(exposureModeIndex);
> +
> +	double gain = estimateInitialGain();
> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> +
> +	/*
> +	 * We don't check whether we're already close to the target, because
> +	 * even if the effective exposure value is the same as the last frame's
> +	 * we could have switched to an exposure mode that would require a new
> +	 * pass through the splitExposure() function.
> +	 */
> +
> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> +					   * exposureModeHelper->maxGain();
> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> +
> +	/*
> +	 * We filter the exposure value to make sure changes are not too jarring
> +	 * from frame to frame.
> +	 */
> +	newExposureValue = filterExposure(newExposureValue);
> +
> +	frameCount_++;
> +	return exposureModeHelper->splitExposure(newExposureValue);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::resetFrameCount()
> + * \brief Reset the frame counter
> + *
> + * This function resets the internal frame counter, which exists to help the
> + * algorithm decide whether it should respond instantly or not. The expectation
> + * is for derived classes to call this function before each camera start call,
> + * either in configure() or queueRequest() if the frame number is zero.
> + */
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> new file mode 100644
> index 00000000..e48dc498
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.h
> @@ -0,0 +1,91 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> + */
> +
> +#pragma once
> +
> +#include <tuple>
> +#include <vector>
> +
> +#include <libcamera/controls.h>
> +
> +#include "libcamera/internal/yaml_parser.h"
> +
> +#include "exposure_mode_helper.h"
> +#include "histogram.h"
> +
> +namespace libcamera {
> +
> +namespace ipa {
> +
> +class AgcMeanLuminance
> +{
> +public:
> +	AgcMeanLuminance();
> +	virtual ~AgcMeanLuminance() = default;

There were a few small comments from Laurent that got lost
 * destructor in cpp
 * code sytel in enum
 * missing line

Aside from that, I think we should merge it in.

Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>

Cheers,
Stefan

> +
> +	struct AgcConstraint {
> +		enum class Bound {
> +			lower = 0,
> +			upper = 1
> +		};
> +		Bound bound;
> +		double qLo;
> +		double qHi;
> +		double yTarget;
> +	};
> +
> +	int parseTuningData(const YamlObject &tuningData);
> +
> +	void configureExposureModeHelpers(utils::Duration minShutter,
> +					  utils::Duration maxShutter,
> +					  double minGain,
> +					  double maxGain);
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> +	{
> +		return constraintModes_;
> +	}
> +
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> +	{
> +		return exposureModeHelpers_;
> +	}
> +
> +	ControlInfoMap::Map controls()
> +	{
> +		return controls_;
> +	}
> +
> +	double estimateInitialGain();
> +	double constraintClampGain(uint32_t constraintModeIndex,
> +				   const Histogram &hist,
> +				   double gain);
> +	utils::Duration filterExposure(utils::Duration exposureValue);
> +	std::tuple<utils::Duration, double, double>
> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> +	void resetFrameCount() { frameCount_ = 0; }
> +private:
> +	virtual double estimateLuminance(const double gain) = 0;
> +
> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> +	int parseConstraintModes(const YamlObject &tuningData);
> +	int parseExposureModes(const YamlObject &tuningData);
> +
> +	uint64_t frameCount_;
> +	utils::Duration filteredExposure_;
> +	double relativeLuminanceTarget_;
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> +	ControlInfoMap::Map controls_;
> +};
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> index 37fbd177..7ce885da 100644
> --- a/src/ipa/libipa/meson.build
> +++ b/src/ipa/libipa/meson.build
> @@ -1,6 +1,7 @@
>  # SPDX-License-Identifier: CC0-1.0
>  
>  libipa_headers = files([
> +    'agc_mean_luminance.h',
>      'algorithm.h',
>      'camera_sensor_helper.h',
>      'exposure_mode_helper.h',
> @@ -10,6 +11,7 @@ libipa_headers = files([
>  ])
>  
>  libipa_sources = files([
> +    'agc_mean_luminance.cpp',
>      'algorithm.cpp',
>      'camera_sensor_helper.cpp',
>      'exposure_mode_helper.cpp',
> -- 
> 2.34.1
>

Hi Dan,

On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> very large part; following the Rpi IPA's algorithm in spirit with a
> few tunable values in that IPA being hardcoded in the libipa ones.
> Add a new base class for MeanLuminanceAgc which implements the same
> algorithm and additionally parses yaml tuning files to inform an IPA
> module's Agc algorithm about valid constraint and exposure modes and
> their associated bounds.
> 
> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> ---
> Changes in v2:
> 
> 	- Renamed the class and files
> 	- Expanded the documentation
> 	- Added parseTuningData() so derived classes can call a single function
> 	  to cover all the parsing in ::init().
> 
>  src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
>  src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
>  src/ipa/libipa/meson.build            |   2 +
>  3 files changed, 674 insertions(+)
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.h
> 
> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> new file mode 100644
> index 00000000..02e223cf
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> @@ -0,0 +1,581 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> + */
> +
> +#include "agc_mean_luminance.h"
> +
> +#include <cmath>
> +
> +#include <libcamera/base/log.h>
> +#include <libcamera/control_ids.h>
> +
> +#include "exposure_mode_helper.h"
> +
> +using namespace libcamera::controls;
> +
> +/**
> + * \file agc_mean_luminance.h
> + * \brief Base class implementing mean luminance AEGC
> + */
> +
> +namespace libcamera {
> +
> +using namespace std::literals::chrono_literals;
> +
> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> +
> +namespace ipa {
> +
> +/*
> + * Number of frames for which to run the algorithm at full speed, before slowing
> + * down to prevent large and jarring changes in exposure from frame to frame.
> + */
> +static constexpr uint32_t kNumStartupFrames = 10;
> +
> +/*
> + * Default relative luminance target
> + *
> + * This value should be chosen so that when the camera points at a grey target,
> + * the resulting image brightness looks "right". Custom values can be passed
> + * as the relativeLuminanceTarget value in sensor tuning files.
> + */
> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> +
> +/**
> + * \struct AgcMeanLuminance::AgcConstraint
> + * \brief The boundaries and target for an AeConstraintMode constraint
> + *
> + * This structure describes an AeConstraintMode constraint for the purposes of
> + * this algorithm. The algorithm will apply the constraints by calculating the
> + * Histogram's inter-quantile mean between the given quantiles and ensure that
> + * the resulting value is the right side of the given target (as defined by the
> + * boundary and luminance target).
> + */
> +
> +/**
> + * \enum AgcMeanLuminance::AgcConstraint::Bound
> + * \brief Specify whether the constraint defines a lower or upper bound
> + * \var AgcMeanLuminance::AgcConstraint::lower
> + * \brief The constraint defines a lower bound
> + * \var AgcMeanLuminance::AgcConstraint::upper
> + * \brief The constraint defines an upper bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::bound
> + * \brief The type of constraint bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qLo
> + * \brief The lower quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qHi
> + * \brief The upper quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::yTarget
> + * \brief The luminance target for the constraint
> + */
> +
> +/**
> + * \class AgcMeanLuminance
> + * \brief A mean-based auto-exposure algorithm
> + *
> + * This algorithm calculates a shutter time, analogue and digital gain such that
> + * the normalised mean luminance value of an image is driven towards a target,
> + * which itself is discovered from tuning data. The algorithm is a two-stage
> + * process.
> + *
> + * In the first stage, an initial gain value is derived by iteratively comparing
> + * the gain-adjusted mean luminance across an entire image against a target, and
> + * selecting a value which pushes it as closely as possible towards the target.
> + *
> + * In the second stage we calculate the gain required to drive the average of a
> + * section of a histogram to a target value, where the target and the boundaries
> + * of the section of the histogram used in the calculation are taken from the
> + * values defined for the currently configured AeConstraintMode within the
> + * tuning data. This class provides a helper function to parse those tuning data
> + * to discover the constraints, and so requires a specific format for those
> + * data which is described in \ref parseTuningData(). The gain from the first
> + * stage is then clamped to the gain from this stage.
> + *
> + * The final gain is used to adjust the effective exposure value of the image,
> + * and that new exposure value is divided into shutter time, analogue gain and
> + * digital gain according to the selected AeExposureMode. This class expects to
> + * use the \ref ExposureModeHelper class to assist in that division, and expects
> + * the data needed to initialise that class to be present in tuning data in a
> + * format described in \ref parseTuningData().
> + *
> + * In order to be able to derive an AGC implementation from this class, an IPA
> + * needs to be able to do the following:
> + *
> + * 1. Provide a luminance estimation across an entire image.
> + * 2. Provide a luminance Histogram for the image to use in calculating
> + *    constraint compliance. The precision of the Histogram that is available
> + *    will determine the supportable precision of the constraints.
> + */
> +
> +AgcMeanLuminance::AgcMeanLuminance()
> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> +{
> +}
> +
> +/**
> + * \brief Parse the relative luminance target from the tuning data
> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> + */
> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> +{
> +	relativeLuminanceTarget_ =
> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> +}
> +
> +/**
> + * \brief Parse an AeConstraintMode constraint from tuning data
> + * \param[in] modeDict the YamlObject holding the constraint data
> + * \param[in] id The constraint ID from AeConstraintModeEnum
> + */
> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> +{
> +	for (const auto &[boundName, content] : modeDict.asDict()) {
> +		if (boundName != "upper" && boundName != "lower") {
> +			LOG(AgcMeanLuminance, Warning)
> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> +			continue;
> +		}
> +
> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> +		double qLo = content["qLo"].get<double>().value_or(0.98);
> +		double qHi = content["qHi"].get<double>().value_or(1.0);
> +		double yTarget =
> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> +
> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> +
> +		if (!constraintModes_.count(id))
> +			constraintModes_[id] = {};
> +
> +		if (idx)
> +			constraintModes_[id].push_back(constraint);
> +		else
> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> +	}
> +}
> +
> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableConstraintModes;
> +
> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> +	if (yamlConstraintModes.isDictionary()) {
> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> +			if (AeConstraintModeNameValueMap.find(modeName) ==
> +			    AeConstraintModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown constraint mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeDict.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid constraint mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			parseConstraint(modeDict,
> +					AeConstraintModeNameValueMap.at(modeName));
> +			availableConstraintModes.push_back(
> +				AeConstraintModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If the tuning data file contains no constraints then we use the
> +	 * default constraint that the various Agc algorithms were adhering to
> +	 * anyway before centralisation.
> +	 */
> +	if (constraintModes_.empty()) {
> +		AgcConstraint constraint = {
> +			AgcConstraint::Bound::lower,
> +			0.98,
> +			1.0,
> +			0.5
> +		};
> +
> +		constraintModes_[controls::ConstraintNormal].insert(
> +			constraintModes_[controls::ConstraintNormal].begin(),
> +			constraint);
> +		availableConstraintModes.push_back(
> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> +	}
> +
> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> +
> +	return 0;
> +}
> +
> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableExposureModes;
> +
> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> +	if (yamlExposureModes.isDictionary()) {
> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> +			if (AeExposureModeNameValueMap.find(modeName) ==
> +			    AeExposureModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown exposure mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeValues.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid exposure mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<uint32_t> shutters =
> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> +			std::vector<double> gains =
> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> +
> +			if (shutters.size() != gains.size()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain array sizes unequal";
> +				return -EINVAL;
> +			}
> +
> +			if (shutters.empty()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain arrays are empty";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<std::pair<utils::Duration, double>> stages;
> +			for (unsigned int i = 0; i < shutters.size(); i++) {
> +				stages.push_back({
> +					std::chrono::microseconds(shutters[i]),
> +					gains[i]
> +				});
> +			}

I was wondering if we could move a significant portion of this to
ExposureModeHelper::readYaml() but since this is the only user (so far)
of it I suppose it doesn't really matter.

Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>

> +
> +			std::shared_ptr<ExposureModeHelper> helper =
> +				std::make_shared<ExposureModeHelper>();
> +			helper->init(stages);
> +
> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If we don't have any exposure modes in the tuning data we create an
> +	 * ExposureModeHelper using an empty vector of stages. This will result
> +	 * in the ExposureModeHelper simply driving the shutter as high as
> +	 * possible before touching gain.
> +	 */
> +	if (availableExposureModes.empty()) {
> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> +		std::vector<std::pair<utils::Duration, double>> stages = { };
> +
> +		std::shared_ptr<ExposureModeHelper> helper =
> +			std::make_shared<ExposureModeHelper>();
> +		helper->init(stages);
> +
> +		exposureModeHelpers_[exposureModeId] = helper;
> +		availableExposureModes.push_back(exposureModeId);
> +	}
> +
> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> + * \param[in] tuningData the YamlObject representing the tuning data
> + *
> + * This function parses tuning data to build the list of allowed values for the
> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> + * the data in a specific format; the Agc algorithm's tuning data should contain
> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> + * Each mode dict may contain either a "lower" or "upper" key or both, for
> + * example:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeConstraintMode:
> + *         ConstraintNormal:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *         ConstraintHighlight:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *           upper:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.8
> + *
> + * \endcode
> + *
> + * For the AeExposureMode control the data should contain a dictionary called
> + * AeExposureMode containing per-mode setting dictionaries with the key being a
> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> + * contain an array of shutter times with the key "shutter" and an array of gain
> + * values with the key "gain", in this format:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeExposureMode:
> + *         ExposureNormal:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *         ExposureShort:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *
> + * \endcode
> + *
> + * @return 0 on success or a negative error code
> + */
> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> +{
> +	int ret;
> +
> +	parseRelativeLuminanceTarget(tuningData);
> +
> +	ret = parseConstraintModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	ret = parseExposureModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief configure the ExposureModeHelpers for this class
> + * \param[in] minShutter Minimum shutter time to allow
> + * \param[in] maxShutter Maximum shutter time to allow
> + * \param[in] minGain Minimum gain to allow
> + * \param[in] maxGain Maximum gain to allow
> + *
> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> + * ExposureModeHelper that has been created for this class.
> + */
> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> +						    utils::Duration maxShutter,
> +						    double minGain,
> +						    double maxGain)
> +{
> +	for (auto &[id, helper] : exposureModeHelpers_)
> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::constraintModes()
> + * \brief Get the constraint modes that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::exposureModeHelpers()
> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::controls()
> + * \brief Get the controls that have been generated after parsing tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> + * \brief Estimate the luminance of an image, adjusted by a given gain
> + * \param[in] gain The gain with which to adjust the luminance estimate
> + *
> + * This function estimates the average relative luminance of the frame that
> + * would be output by the sensor if an additional \a gain was applied. It is a
> + * pure virtual function because estimation of luminance is a hardware-specific
> + * operation, which depends wholly on the format of the stats that are delivered
> + * to libcamera from the ISP. Derived classes must implement an overriding
> + * function that calculates the normalised mean luminance value across the
> + * entire image.
> + *
> + * \return The normalised relative luminance of the image
> + */
> +
> +/**
> + * \brief Estimate the initial gain needed to achieve a relative luminance
> + * target
> + *
> + * To account for non-linearity caused by saturation, the value needs to be
> + * estimated in an iterative process, as multiplying by a gain will not increase
> + * the relative luminance by the same factor if some image regions are saturated
> + *
> + * \return The calculated initial gain
> + */
> +double AgcMeanLuminance::estimateInitialGain()
> +{
> +	double yTarget = relativeLuminanceTarget_;
> +	double yGain = 1.0;
> +
> +	for (unsigned int i = 0; i < 8; i++) {
> +		double yValue = estimateLuminance(yGain);
> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> +
> +		yGain *= extra_gain;
> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> +				<< ", Y target: " << yTarget
> +				<< ", gives gain " << yGain;
> +
> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> +			break;
> +	}
> +
> +	return yGain;
> +}
> +
> +/**
> + * \brief Clamp gain within the bounds of a defined constraint
> + * \param[in] constraintModeIndex The index of the constraint to adhere to
> + * \param[in] hist A histogram over which to calculate inter-quantile means
> + * \param[in] gain The gain to clamp
> + *
> + * \return The gain clamped within the constraint bounds
> + */
> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> +					     const Histogram &hist,
> +					     double gain)
> +{
> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> +	for (const AgcConstraint &constraint : constraints) {
> +		double newGain = constraint.yTarget * hist.bins() /
> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> +
> +		if (constraint.bound == AgcConstraint::Bound::lower &&
> +		    newGain > gain)
> +			gain = newGain;
> +
> +		if (constraint.bound == AgcConstraint::Bound::upper &&
> +		    newGain < gain)
> +			gain = newGain;
> +	}
> +
> +	return gain;
> +}
> +
> +/**
> + * \brief Apply a filter on the exposure value to limit the speed of changes
> + * \param[in] exposureValue The target exposure from the AGC algorithm
> + *
> + * The speed of the filter is adaptive, and will produce the target quicker
> + * during startup, or when the target exposure is within 20% of the most recent
> + * filter output.
> + *
> + * \return The filtered exposure
> + */
> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> +{
> +	double speed = 0.2;
> +
> +	/* Adapt instantly if we are in startup phase. */
> +	if (frameCount_ < kNumStartupFrames)
> +		speed = 1.0;
> +
> +	/*
> +	 * If we are close to the desired result, go faster to avoid making
> +	 * multiple micro-adjustments.
> +	 * \todo Make this customisable?
> +	 */
> +	if (filteredExposure_ < 1.2 * exposureValue &&
> +	    filteredExposure_ > 0.8 * exposureValue)
> +		speed = sqrt(speed);
> +
> +	filteredExposure_ = speed * exposureValue +
> +			    filteredExposure_ * (1.0 - speed);
> +
> +	return filteredExposure_;
> +}
> +
> +/**
> + * \brief Calculate the new exposure value
> + * \param[in] constraintModeIndex The index of the current constraint mode
> + * \param[in] exposureModeIndex The index of the current exposure mode
> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> + *	      the calculated gain
> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> + *	      statistics in use derive
> + *
> + * Calculate a new exposure value to try to obtain the target. The calculated
> + * exposure value is filtered to prevent rapid changes from frame to frame, and
> + * divided into shutter time, analogue and digital gain.
> + *
> + * \return Tuple of shutter time, analogue gain, and digital gain
> + */
> +std::tuple<utils::Duration, double, double>
> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> +				 uint32_t exposureModeIndex,
> +				 const Histogram &yHist,
> +				 utils::Duration effectiveExposureValue)
> +{
> +	/*
> +	 * The pipeline handler should validate that we have received an allowed
> +	 * value for AeExposureMode.
> +	 */
> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> +		exposureModeHelpers_.at(exposureModeIndex);
> +
> +	double gain = estimateInitialGain();
> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> +
> +	/*
> +	 * We don't check whether we're already close to the target, because
> +	 * even if the effective exposure value is the same as the last frame's
> +	 * we could have switched to an exposure mode that would require a new
> +	 * pass through the splitExposure() function.
> +	 */
> +
> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> +					   * exposureModeHelper->maxGain();
> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> +
> +	/*
> +	 * We filter the exposure value to make sure changes are not too jarring
> +	 * from frame to frame.
> +	 */
> +	newExposureValue = filterExposure(newExposureValue);
> +
> +	frameCount_++;
> +	return exposureModeHelper->splitExposure(newExposureValue);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::resetFrameCount()
> + * \brief Reset the frame counter
> + *
> + * This function resets the internal frame counter, which exists to help the
> + * algorithm decide whether it should respond instantly or not. The expectation
> + * is for derived classes to call this function before each camera start call,
> + * either in configure() or queueRequest() if the frame number is zero.
> + */
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> new file mode 100644
> index 00000000..e48dc498
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.h
> @@ -0,0 +1,91 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> + */
> +
> +#pragma once
> +
> +#include <tuple>
> +#include <vector>
> +
> +#include <libcamera/controls.h>
> +
> +#include "libcamera/internal/yaml_parser.h"
> +
> +#include "exposure_mode_helper.h"
> +#include "histogram.h"
> +
> +namespace libcamera {
> +
> +namespace ipa {
> +
> +class AgcMeanLuminance
> +{
> +public:
> +	AgcMeanLuminance();
> +	virtual ~AgcMeanLuminance() = default;
> +
> +	struct AgcConstraint {
> +		enum class Bound {
> +			lower = 0,
> +			upper = 1
> +		};
> +		Bound bound;
> +		double qLo;
> +		double qHi;
> +		double yTarget;
> +	};
> +
> +	int parseTuningData(const YamlObject &tuningData);
> +
> +	void configureExposureModeHelpers(utils::Duration minShutter,
> +					  utils::Duration maxShutter,
> +					  double minGain,
> +					  double maxGain);
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> +	{
> +		return constraintModes_;
> +	}
> +
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> +	{
> +		return exposureModeHelpers_;
> +	}
> +
> +	ControlInfoMap::Map controls()
> +	{
> +		return controls_;
> +	}
> +
> +	double estimateInitialGain();
> +	double constraintClampGain(uint32_t constraintModeIndex,
> +				   const Histogram &hist,
> +				   double gain);
> +	utils::Duration filterExposure(utils::Duration exposureValue);
> +	std::tuple<utils::Duration, double, double>
> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> +	void resetFrameCount() { frameCount_ = 0; }
> +private:
> +	virtual double estimateLuminance(const double gain) = 0;
> +
> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> +	int parseConstraintModes(const YamlObject &tuningData);
> +	int parseExposureModes(const YamlObject &tuningData);
> +
> +	uint64_t frameCount_;
> +	utils::Duration filteredExposure_;
> +	double relativeLuminanceTarget_;
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> +	ControlInfoMap::Map controls_;
> +};
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> index 37fbd177..7ce885da 100644
> --- a/src/ipa/libipa/meson.build
> +++ b/src/ipa/libipa/meson.build
> @@ -1,6 +1,7 @@
>  # SPDX-License-Identifier: CC0-1.0
>  
>  libipa_headers = files([
> +    'agc_mean_luminance.h',
>      'algorithm.h',
>      'camera_sensor_helper.h',
>      'exposure_mode_helper.h',
> @@ -10,6 +11,7 @@ libipa_headers = files([
>  ])
>  
>  libipa_sources = files([
> +    'agc_mean_luminance.cpp',
>      'algorithm.cpp',
>      'camera_sensor_helper.cpp',
>      'exposure_mode_helper.cpp',
> -- 
> 2.34.1
>

Hi Paul

On 17/04/2024 14:15, Daniel Scally wrote:
> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> very large part; following the Rpi IPA's algorithm in spirit with a
> few tunable values in that IPA being hardcoded in the libipa ones.
> Add a new base class for MeanLuminanceAgc which implements the same
> algorithm and additionally parses yaml tuning files to inform an IPA
> module's Agc algorithm about valid constraint and exposure modes and
> their associated bounds.
>
> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> ---
> Changes in v2:
>
> 	- Renamed the class and files
> 	- Expanded the documentation
> 	- Added parseTuningData() so derived classes can call a single function
> 	  to cover all the parsing in ::init().
>
>   src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
>   src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
>   src/ipa/libipa/meson.build            |   2 +
>   3 files changed, 674 insertions(+)
>   create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
>   create mode 100644 src/ipa/libipa/agc_mean_luminance.h
>
> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> new file mode 100644
> index 00000000..02e223cf
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> @@ -0,0 +1,581 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> + */
> +
> +#include "agc_mean_luminance.h"
> +
> +#include <cmath>
> +
> +#include <libcamera/base/log.h>
> +#include <libcamera/control_ids.h>
> +
> +#include "exposure_mode_helper.h"
> +
> +using namespace libcamera::controls;
> +
> +/**
> + * \file agc_mean_luminance.h
> + * \brief Base class implementing mean luminance AEGC
> + */
> +
> +namespace libcamera {
> +
> +using namespace std::literals::chrono_literals;
> +
> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> +
> +namespace ipa {
> +
> +/*
> + * Number of frames for which to run the algorithm at full speed, before slowing
> + * down to prevent large and jarring changes in exposure from frame to frame.
> + */
> +static constexpr uint32_t kNumStartupFrames = 10;
> +
> +/*
> + * Default relative luminance target
> + *
> + * This value should be chosen so that when the camera points at a grey target,
> + * the resulting image brightness looks "right". Custom values can be passed
> + * as the relativeLuminanceTarget value in sensor tuning files.
> + */
> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> +
> +/**
> + * \struct AgcMeanLuminance::AgcConstraint
> + * \brief The boundaries and target for an AeConstraintMode constraint
> + *
> + * This structure describes an AeConstraintMode constraint for the purposes of
> + * this algorithm. The algorithm will apply the constraints by calculating the
> + * Histogram's inter-quantile mean between the given quantiles and ensure that
> + * the resulting value is the right side of the given target (as defined by the
> + * boundary and luminance target).
> + */
> +
> +/**
> + * \enum AgcMeanLuminance::AgcConstraint::Bound
> + * \brief Specify whether the constraint defines a lower or upper bound
> + * \var AgcMeanLuminance::AgcConstraint::lower
> + * \brief The constraint defines a lower bound
> + * \var AgcMeanLuminance::AgcConstraint::upper
> + * \brief The constraint defines an upper bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::bound
> + * \brief The type of constraint bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qLo
> + * \brief The lower quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qHi
> + * \brief The upper quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::yTarget
> + * \brief The luminance target for the constraint
> + */
> +
> +/**
> + * \class AgcMeanLuminance
> + * \brief A mean-based auto-exposure algorithm
> + *
> + * This algorithm calculates a shutter time, analogue and digital gain such that
> + * the normalised mean luminance value of an image is driven towards a target,
> + * which itself is discovered from tuning data. The algorithm is a two-stage
> + * process.
> + *
> + * In the first stage, an initial gain value is derived by iteratively comparing
> + * the gain-adjusted mean luminance across an entire image against a target, and
> + * selecting a value which pushes it as closely as possible towards the target.
> + *
> + * In the second stage we calculate the gain required to drive the average of a
> + * section of a histogram to a target value, where the target and the boundaries
> + * of the section of the histogram used in the calculation are taken from the
> + * values defined for the currently configured AeConstraintMode within the
> + * tuning data. This class provides a helper function to parse those tuning data
> + * to discover the constraints, and so requires a specific format for those
> + * data which is described in \ref parseTuningData(). The gain from the first
> + * stage is then clamped to the gain from this stage.
> + *
> + * The final gain is used to adjust the effective exposure value of the image,
> + * and that new exposure value is divided into shutter time, analogue gain and
> + * digital gain according to the selected AeExposureMode. This class expects to
> + * use the \ref ExposureModeHelper class to assist in that division, and expects
> + * the data needed to initialise that class to be present in tuning data in a
> + * format described in \ref parseTuningData().
> + *
> + * In order to be able to derive an AGC implementation from this class, an IPA
> + * needs to be able to do the following:
> + *
> + * 1. Provide a luminance estimation across an entire image.
> + * 2. Provide a luminance Histogram for the image to use in calculating
> + *    constraint compliance. The precision of the Histogram that is available
> + *    will determine the supportable precision of the constraints.
> + */
> +
> +AgcMeanLuminance::AgcMeanLuminance()
> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> +{
> +}
> +
> +/**
> + * \brief Parse the relative luminance target from the tuning data
> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> + */
> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> +{
> +	relativeLuminanceTarget_ =
> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> +}
> +
> +/**
> + * \brief Parse an AeConstraintMode constraint from tuning data
> + * \param[in] modeDict the YamlObject holding the constraint data
> + * \param[in] id The constraint ID from AeConstraintModeEnum
> + */
> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> +{
> +	for (const auto &[boundName, content] : modeDict.asDict()) {
> +		if (boundName != "upper" && boundName != "lower") {
> +			LOG(AgcMeanLuminance, Warning)
> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> +			continue;
> +		}
> +
> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> +		double qLo = content["qLo"].get<double>().value_or(0.98);
> +		double qHi = content["qHi"].get<double>().value_or(1.0);
> +		double yTarget =
> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> +


Sorry, I did see your point in the last thread that this should be a piecewise linear function but I 
forgot to respond to it. I do wonder if it's actually necessary for this one to be a piecewise 
linear function since, although Rpi treats it as a PWL, the points they define in their tuning data 
are such that it's actually a static value whatever the lux...but perhaps it's better to be 
consistent in how we treat it.

> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> +
> +		if (!constraintModes_.count(id))
> +			constraintModes_[id] = {};
> +
> +		if (idx)
> +			constraintModes_[id].push_back(constraint);
> +		else
> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> +	}
> +}
> +
> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableConstraintModes;
> +
> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> +	if (yamlConstraintModes.isDictionary()) {
> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> +			if (AeConstraintModeNameValueMap.find(modeName) ==
> +			    AeConstraintModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown constraint mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeDict.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid constraint mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			parseConstraint(modeDict,
> +					AeConstraintModeNameValueMap.at(modeName));
> +			availableConstraintModes.push_back(
> +				AeConstraintModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If the tuning data file contains no constraints then we use the
> +	 * default constraint that the various Agc algorithms were adhering to
> +	 * anyway before centralisation.
> +	 */
> +	if (constraintModes_.empty()) {
> +		AgcConstraint constraint = {
> +			AgcConstraint::Bound::lower,
> +			0.98,
> +			1.0,
> +			0.5
> +		};
> +
> +		constraintModes_[controls::ConstraintNormal].insert(
> +			constraintModes_[controls::ConstraintNormal].begin(),
> +			constraint);
> +		availableConstraintModes.push_back(
> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> +	}
> +
> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> +
> +	return 0;
> +}
> +
> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableExposureModes;
> +
> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> +	if (yamlExposureModes.isDictionary()) {
> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> +			if (AeExposureModeNameValueMap.find(modeName) ==
> +			    AeExposureModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown exposure mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeValues.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid exposure mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<uint32_t> shutters =
> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> +			std::vector<double> gains =
> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> +
> +			if (shutters.size() != gains.size()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain array sizes unequal";
> +				return -EINVAL;
> +			}
> +
> +			if (shutters.empty()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain arrays are empty";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<std::pair<utils::Duration, double>> stages;
> +			for (unsigned int i = 0; i < shutters.size(); i++) {
> +				stages.push_back({
> +					std::chrono::microseconds(shutters[i]),
> +					gains[i]
> +				});
> +			}
> +
> +			std::shared_ptr<ExposureModeHelper> helper =
> +				std::make_shared<ExposureModeHelper>();
> +			helper->init(stages);
> +
> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If we don't have any exposure modes in the tuning data we create an
> +	 * ExposureModeHelper using an empty vector of stages. This will result
> +	 * in the ExposureModeHelper simply driving the shutter as high as
> +	 * possible before touching gain.
> +	 */
> +	if (availableExposureModes.empty()) {
> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> +		std::vector<std::pair<utils::Duration, double>> stages = { };
> +
> +		std::shared_ptr<ExposureModeHelper> helper =
> +			std::make_shared<ExposureModeHelper>();
> +		helper->init(stages);
> +
> +		exposureModeHelpers_[exposureModeId] = helper;
> +		availableExposureModes.push_back(exposureModeId);
> +	}
> +
> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> + * \param[in] tuningData the YamlObject representing the tuning data
> + *
> + * This function parses tuning data to build the list of allowed values for the
> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> + * the data in a specific format; the Agc algorithm's tuning data should contain
> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> + * Each mode dict may contain either a "lower" or "upper" key or both, for
> + * example:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeConstraintMode:
> + *         ConstraintNormal:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *         ConstraintHighlight:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *           upper:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.8
> + *
> + * \endcode
> + *
> + * For the AeExposureMode control the data should contain a dictionary called
> + * AeExposureMode containing per-mode setting dictionaries with the key being a
> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> + * contain an array of shutter times with the key "shutter" and an array of gain
> + * values with the key "gain", in this format:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeExposureMode:
> + *         ExposureNormal:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *         ExposureShort:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *
> + * \endcode
> + *
> + * @return 0 on success or a negative error code
> + */
> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> +{
> +	int ret;
> +
> +	parseRelativeLuminanceTarget(tuningData);
> +
> +	ret = parseConstraintModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	ret = parseExposureModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief configure the ExposureModeHelpers for this class
> + * \param[in] minShutter Minimum shutter time to allow
> + * \param[in] maxShutter Maximum shutter time to allow
> + * \param[in] minGain Minimum gain to allow
> + * \param[in] maxGain Maximum gain to allow
> + *
> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> + * ExposureModeHelper that has been created for this class.
> + */
> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> +						    utils::Duration maxShutter,
> +						    double minGain,
> +						    double maxGain)
> +{
> +	for (auto &[id, helper] : exposureModeHelpers_)
> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::constraintModes()
> + * \brief Get the constraint modes that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::exposureModeHelpers()
> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::controls()
> + * \brief Get the controls that have been generated after parsing tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> + * \brief Estimate the luminance of an image, adjusted by a given gain
> + * \param[in] gain The gain with which to adjust the luminance estimate
> + *
> + * This function estimates the average relative luminance of the frame that
> + * would be output by the sensor if an additional \a gain was applied. It is a
> + * pure virtual function because estimation of luminance is a hardware-specific
> + * operation, which depends wholly on the format of the stats that are delivered
> + * to libcamera from the ISP. Derived classes must implement an overriding
> + * function that calculates the normalised mean luminance value across the
> + * entire image.
> + *
> + * \return The normalised relative luminance of the image
> + */
> +
> +/**
> + * \brief Estimate the initial gain needed to achieve a relative luminance
> + * target
> + *
> + * To account for non-linearity caused by saturation, the value needs to be
> + * estimated in an iterative process, as multiplying by a gain will not increase
> + * the relative luminance by the same factor if some image regions are saturated
> + *
> + * \return The calculated initial gain
> + */
> +double AgcMeanLuminance::estimateInitialGain()
> +{
> +	double yTarget = relativeLuminanceTarget_;
> +	double yGain = 1.0;
> +
> +	for (unsigned int i = 0; i < 8; i++) {
> +		double yValue = estimateLuminance(yGain);
> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> +
> +		yGain *= extra_gain;
> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> +				<< ", Y target: " << yTarget
> +				<< ", gives gain " << yGain;
> +
> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> +			break;
> +	}
> +
> +	return yGain;
> +}
> +
> +/**
> + * \brief Clamp gain within the bounds of a defined constraint
> + * \param[in] constraintModeIndex The index of the constraint to adhere to
> + * \param[in] hist A histogram over which to calculate inter-quantile means
> + * \param[in] gain The gain to clamp
> + *
> + * \return The gain clamped within the constraint bounds
> + */
> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> +					     const Histogram &hist,
> +					     double gain)
> +{
> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> +	for (const AgcConstraint &constraint : constraints) {
> +		double newGain = constraint.yTarget * hist.bins() /
> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> +
> +		if (constraint.bound == AgcConstraint::Bound::lower &&
> +		    newGain > gain)
> +			gain = newGain;
> +
> +		if (constraint.bound == AgcConstraint::Bound::upper &&
> +		    newGain < gain)
> +			gain = newGain;
> +	}
> +
> +	return gain;
> +}
> +
> +/**
> + * \brief Apply a filter on the exposure value to limit the speed of changes
> + * \param[in] exposureValue The target exposure from the AGC algorithm
> + *
> + * The speed of the filter is adaptive, and will produce the target quicker
> + * during startup, or when the target exposure is within 20% of the most recent
> + * filter output.
> + *
> + * \return The filtered exposure
> + */
> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> +{
> +	double speed = 0.2;
> +
> +	/* Adapt instantly if we are in startup phase. */
> +	if (frameCount_ < kNumStartupFrames)
> +		speed = 1.0;
> +
> +	/*
> +	 * If we are close to the desired result, go faster to avoid making
> +	 * multiple micro-adjustments.
> +	 * \todo Make this customisable?
> +	 */
> +	if (filteredExposure_ < 1.2 * exposureValue &&
> +	    filteredExposure_ > 0.8 * exposureValue)
> +		speed = sqrt(speed);
> +
> +	filteredExposure_ = speed * exposureValue +
> +			    filteredExposure_ * (1.0 - speed);
> +
> +	return filteredExposure_;
> +}
> +
> +/**
> + * \brief Calculate the new exposure value
> + * \param[in] constraintModeIndex The index of the current constraint mode
> + * \param[in] exposureModeIndex The index of the current exposure mode
> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> + *	      the calculated gain
> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> + *	      statistics in use derive
> + *
> + * Calculate a new exposure value to try to obtain the target. The calculated
> + * exposure value is filtered to prevent rapid changes from frame to frame, and
> + * divided into shutter time, analogue and digital gain.
> + *
> + * \return Tuple of shutter time, analogue gain, and digital gain
> + */
> +std::tuple<utils::Duration, double, double>
> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> +				 uint32_t exposureModeIndex,
> +				 const Histogram &yHist,
> +				 utils::Duration effectiveExposureValue)
> +{
> +	/*
> +	 * The pipeline handler should validate that we have received an allowed
> +	 * value for AeExposureMode.
> +	 */
> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> +		exposureModeHelpers_.at(exposureModeIndex);
> +
> +	double gain = estimateInitialGain();
> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> +
> +	/*
> +	 * We don't check whether we're already close to the target, because
> +	 * even if the effective exposure value is the same as the last frame's
> +	 * we could have switched to an exposure mode that would require a new
> +	 * pass through the splitExposure() function.
> +	 */
> +
> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> +					   * exposureModeHelper->maxGain();
> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> +
> +	/*
> +	 * We filter the exposure value to make sure changes are not too jarring
> +	 * from frame to frame.
> +	 */
> +	newExposureValue = filterExposure(newExposureValue);
> +
> +	frameCount_++;
> +	return exposureModeHelper->splitExposure(newExposureValue);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::resetFrameCount()
> + * \brief Reset the frame counter
> + *
> + * This function resets the internal frame counter, which exists to help the
> + * algorithm decide whether it should respond instantly or not. The expectation
> + * is for derived classes to call this function before each camera start call,
> + * either in configure() or queueRequest() if the frame number is zero.
> + */
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> new file mode 100644
> index 00000000..e48dc498
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.h
> @@ -0,0 +1,91 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> + */
> +
> +#pragma once
> +
> +#include <tuple>
> +#include <vector>
> +
> +#include <libcamera/controls.h>
> +
> +#include "libcamera/internal/yaml_parser.h"
> +
> +#include "exposure_mode_helper.h"
> +#include "histogram.h"
> +
> +namespace libcamera {
> +
> +namespace ipa {
> +
> +class AgcMeanLuminance
> +{
> +public:
> +	AgcMeanLuminance();
> +	virtual ~AgcMeanLuminance() = default;
> +
> +	struct AgcConstraint {
> +		enum class Bound {
> +			lower = 0,
> +			upper = 1
> +		};
> +		Bound bound;
> +		double qLo;
> +		double qHi;
> +		double yTarget;
> +	};
> +
> +	int parseTuningData(const YamlObject &tuningData);
> +
> +	void configureExposureModeHelpers(utils::Duration minShutter,
> +					  utils::Duration maxShutter,
> +					  double minGain,
> +					  double maxGain);
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> +	{
> +		return constraintModes_;
> +	}
> +
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> +	{
> +		return exposureModeHelpers_;
> +	}
> +
> +	ControlInfoMap::Map controls()
> +	{
> +		return controls_;
> +	}
> +
> +	double estimateInitialGain();
> +	double constraintClampGain(uint32_t constraintModeIndex,
> +				   const Histogram &hist,
> +				   double gain);
> +	utils::Duration filterExposure(utils::Duration exposureValue);
> +	std::tuple<utils::Duration, double, double>
> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> +	void resetFrameCount() { frameCount_ = 0; }
> +private:
> +	virtual double estimateLuminance(const double gain) = 0;
> +
> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> +	int parseConstraintModes(const YamlObject &tuningData);
> +	int parseExposureModes(const YamlObject &tuningData);
> +
> +	uint64_t frameCount_;
> +	utils::Duration filteredExposure_;
> +	double relativeLuminanceTarget_;
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> +	ControlInfoMap::Map controls_;
> +};
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> index 37fbd177..7ce885da 100644
> --- a/src/ipa/libipa/meson.build
> +++ b/src/ipa/libipa/meson.build
> @@ -1,6 +1,7 @@
>   # SPDX-License-Identifier: CC0-1.0
>   
>   libipa_headers = files([
> +    'agc_mean_luminance.h',
>       'algorithm.h',
>       'camera_sensor_helper.h',
>       'exposure_mode_helper.h',
> @@ -10,6 +11,7 @@ libipa_headers = files([
>   ])
>   
>   libipa_sources = files([
> +    'agc_mean_luminance.cpp',
>       'algorithm.cpp',
>       'camera_sensor_helper.cpp',
>       'exposure_mode_helper.cpp',

It just occurred to me that the commit message still says
MeanLuminanceAgc instead of AgcMeanLuminance.


Paul

On Thu, Apr 18, 2024 at 09:49:34PM +0900, Paul Elder wrote:
> Hi Dan,
> 
> On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
> > The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> > very large part; following the Rpi IPA's algorithm in spirit with a
> > few tunable values in that IPA being hardcoded in the libipa ones.
> > Add a new base class for MeanLuminanceAgc which implements the same
> > algorithm and additionally parses yaml tuning files to inform an IPA
> > module's Agc algorithm about valid constraint and exposure modes and
> > their associated bounds.
> > 
> > Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> > ---
> > Changes in v2:
> > 
> > 	- Renamed the class and files
> > 	- Expanded the documentation
> > 	- Added parseTuningData() so derived classes can call a single function
> > 	  to cover all the parsing in ::init().
> > 
> >  src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
> >  src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
> >  src/ipa/libipa/meson.build            |   2 +
> >  3 files changed, 674 insertions(+)
> >  create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
> >  create mode 100644 src/ipa/libipa/agc_mean_luminance.h
> > 
> > diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> > new file mode 100644
> > index 00000000..02e223cf
> > --- /dev/null
> > +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> > @@ -0,0 +1,581 @@
> > +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> > +/*
> > + * Copyright (C) 2024 Ideas on Board Oy
> > + *
> > + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> > + */
> > +
> > +#include "agc_mean_luminance.h"
> > +
> > +#include <cmath>
> > +
> > +#include <libcamera/base/log.h>
> > +#include <libcamera/control_ids.h>
> > +
> > +#include "exposure_mode_helper.h"
> > +
> > +using namespace libcamera::controls;
> > +
> > +/**
> > + * \file agc_mean_luminance.h
> > + * \brief Base class implementing mean luminance AEGC
> > + */
> > +
> > +namespace libcamera {
> > +
> > +using namespace std::literals::chrono_literals;
> > +
> > +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> > +
> > +namespace ipa {
> > +
> > +/*
> > + * Number of frames for which to run the algorithm at full speed, before slowing
> > + * down to prevent large and jarring changes in exposure from frame to frame.
> > + */
> > +static constexpr uint32_t kNumStartupFrames = 10;
> > +
> > +/*
> > + * Default relative luminance target
> > + *
> > + * This value should be chosen so that when the camera points at a grey target,
> > + * the resulting image brightness looks "right". Custom values can be passed
> > + * as the relativeLuminanceTarget value in sensor tuning files.
> > + */
> > +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> > +
> > +/**
> > + * \struct AgcMeanLuminance::AgcConstraint
> > + * \brief The boundaries and target for an AeConstraintMode constraint
> > + *
> > + * This structure describes an AeConstraintMode constraint for the purposes of
> > + * this algorithm. The algorithm will apply the constraints by calculating the
> > + * Histogram's inter-quantile mean between the given quantiles and ensure that
> > + * the resulting value is the right side of the given target (as defined by the
> > + * boundary and luminance target).
> > + */
> > +
> > +/**
> > + * \enum AgcMeanLuminance::AgcConstraint::Bound
> > + * \brief Specify whether the constraint defines a lower or upper bound
> > + * \var AgcMeanLuminance::AgcConstraint::lower
> > + * \brief The constraint defines a lower bound
> > + * \var AgcMeanLuminance::AgcConstraint::upper
> > + * \brief The constraint defines an upper bound
> > + */
> > +
> > +/**
> > + * \var AgcMeanLuminance::AgcConstraint::bound
> > + * \brief The type of constraint bound
> > + */
> > +
> > +/**
> > + * \var AgcMeanLuminance::AgcConstraint::qLo
> > + * \brief The lower quantile to use for the constraint
> > + */
> > +
> > +/**
> > + * \var AgcMeanLuminance::AgcConstraint::qHi
> > + * \brief The upper quantile to use for the constraint
> > + */
> > +
> > +/**
> > + * \var AgcMeanLuminance::AgcConstraint::yTarget
> > + * \brief The luminance target for the constraint
> > + */
> > +
> > +/**
> > + * \class AgcMeanLuminance
> > + * \brief A mean-based auto-exposure algorithm
> > + *
> > + * This algorithm calculates a shutter time, analogue and digital gain such that
> > + * the normalised mean luminance value of an image is driven towards a target,
> > + * which itself is discovered from tuning data. The algorithm is a two-stage
> > + * process.
> > + *
> > + * In the first stage, an initial gain value is derived by iteratively comparing
> > + * the gain-adjusted mean luminance across an entire image against a target, and
> > + * selecting a value which pushes it as closely as possible towards the target.
> > + *
> > + * In the second stage we calculate the gain required to drive the average of a
> > + * section of a histogram to a target value, where the target and the boundaries
> > + * of the section of the histogram used in the calculation are taken from the
> > + * values defined for the currently configured AeConstraintMode within the
> > + * tuning data. This class provides a helper function to parse those tuning data
> > + * to discover the constraints, and so requires a specific format for those
> > + * data which is described in \ref parseTuningData(). The gain from the first
> > + * stage is then clamped to the gain from this stage.
> > + *
> > + * The final gain is used to adjust the effective exposure value of the image,
> > + * and that new exposure value is divided into shutter time, analogue gain and
> > + * digital gain according to the selected AeExposureMode. This class expects to
> > + * use the \ref ExposureModeHelper class to assist in that division, and expects
> > + * the data needed to initialise that class to be present in tuning data in a
> > + * format described in \ref parseTuningData().
> > + *
> > + * In order to be able to derive an AGC implementation from this class, an IPA
> > + * needs to be able to do the following:
> > + *
> > + * 1. Provide a luminance estimation across an entire image.
> > + * 2. Provide a luminance Histogram for the image to use in calculating
> > + *    constraint compliance. The precision of the Histogram that is available
> > + *    will determine the supportable precision of the constraints.
> > + */
> > +
> > +AgcMeanLuminance::AgcMeanLuminance()
> > +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> > +{
> > +}
> > +
> > +/**
> > + * \brief Parse the relative luminance target from the tuning data
> > + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> > + */
> > +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> > +{
> > +	relativeLuminanceTarget_ =
> > +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> > +}
> > +
> > +/**
> > + * \brief Parse an AeConstraintMode constraint from tuning data
> > + * \param[in] modeDict the YamlObject holding the constraint data
> > + * \param[in] id The constraint ID from AeConstraintModeEnum
> > + */
> > +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> > +{
> > +	for (const auto &[boundName, content] : modeDict.asDict()) {
> > +		if (boundName != "upper" && boundName != "lower") {
> > +			LOG(AgcMeanLuminance, Warning)
> > +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> > +			continue;
> > +		}
> > +
> > +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> > +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> > +		double qLo = content["qLo"].get<double>().value_or(0.98);
> > +		double qHi = content["qHi"].get<double>().value_or(1.0);
> > +		double yTarget =
> > +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> > +
> > +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> > +
> > +		if (!constraintModes_.count(id))
> > +			constraintModes_[id] = {};
> > +
> > +		if (idx)
> > +			constraintModes_[id].push_back(constraint);
> > +		else
> > +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> > +	}
> > +}
> > +
> > +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> > +{
> > +	std::vector<ControlValue> availableConstraintModes;
> > +
> > +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> > +	if (yamlConstraintModes.isDictionary()) {
> > +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> > +			if (AeConstraintModeNameValueMap.find(modeName) ==
> > +			    AeConstraintModeNameValueMap.end()) {
> > +				LOG(AgcMeanLuminance, Warning)
> > +					<< "Skipping unknown constraint mode '" << modeName << "'";
> > +				continue;
> > +			}
> > +
> > +			if (!modeDict.isDictionary()) {
> > +				LOG(AgcMeanLuminance, Error)
> > +					<< "Invalid constraint mode '" << modeName << "'";
> > +				return -EINVAL;
> > +			}
> > +
> > +			parseConstraint(modeDict,
> > +					AeConstraintModeNameValueMap.at(modeName));
> > +			availableConstraintModes.push_back(
> > +				AeConstraintModeNameValueMap.at(modeName));
> > +		}
> > +	}
> > +
> > +	/*
> > +	 * If the tuning data file contains no constraints then we use the
> > +	 * default constraint that the various Agc algorithms were adhering to
> > +	 * anyway before centralisation.
> > +	 */
> > +	if (constraintModes_.empty()) {
> > +		AgcConstraint constraint = {
> > +			AgcConstraint::Bound::lower,
> > +			0.98,
> > +			1.0,
> > +			0.5
> > +		};
> > +
> > +		constraintModes_[controls::ConstraintNormal].insert(
> > +			constraintModes_[controls::ConstraintNormal].begin(),
> > +			constraint);
> > +		availableConstraintModes.push_back(
> > +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> > +	}
> > +
> > +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> > +
> > +	return 0;
> > +}
> > +
> > +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> > +{
> > +	std::vector<ControlValue> availableExposureModes;
> > +
> > +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> > +	if (yamlExposureModes.isDictionary()) {
> > +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> > +			if (AeExposureModeNameValueMap.find(modeName) ==
> > +			    AeExposureModeNameValueMap.end()) {
> > +				LOG(AgcMeanLuminance, Warning)
> > +					<< "Skipping unknown exposure mode '" << modeName << "'";
> > +				continue;
> > +			}
> > +
> > +			if (!modeValues.isDictionary()) {
> > +				LOG(AgcMeanLuminance, Error)
> > +					<< "Invalid exposure mode '" << modeName << "'";
> > +				return -EINVAL;
> > +			}
> > +
> > +			std::vector<uint32_t> shutters =
> > +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> > +			std::vector<double> gains =
> > +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> > +
> > +			if (shutters.size() != gains.size()) {
> > +				LOG(AgcMeanLuminance, Error)
> > +					<< "Shutter and gain array sizes unequal";
> > +				return -EINVAL;
> > +			}
> > +
> > +			if (shutters.empty()) {
> > +				LOG(AgcMeanLuminance, Error)
> > +					<< "Shutter and gain arrays are empty";
> > +				return -EINVAL;
> > +			}
> > +
> > +			std::vector<std::pair<utils::Duration, double>> stages;
> > +			for (unsigned int i = 0; i < shutters.size(); i++) {
> > +				stages.push_back({
> > +					std::chrono::microseconds(shutters[i]),
> > +					gains[i]
> > +				});
> > +			}
> 
> I was wondering if we could move a significant portion of this to
> ExposureModeHelper::readYaml() but since this is the only user (so far)
> of it I suppose it doesn't really matter.
> 
> Reviewed-by: Paul Elder <paul.elder@ideasonboard.com>
> 
> > +
> > +			std::shared_ptr<ExposureModeHelper> helper =
> > +				std::make_shared<ExposureModeHelper>();
> > +			helper->init(stages);
> > +
> > +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> > +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> > +		}
> > +	}
> > +
> > +	/*
> > +	 * If we don't have any exposure modes in the tuning data we create an
> > +	 * ExposureModeHelper using an empty vector of stages. This will result
> > +	 * in the ExposureModeHelper simply driving the shutter as high as
> > +	 * possible before touching gain.
> > +	 */
> > +	if (availableExposureModes.empty()) {
> > +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> > +		std::vector<std::pair<utils::Duration, double>> stages = { };
> > +
> > +		std::shared_ptr<ExposureModeHelper> helper =
> > +			std::make_shared<ExposureModeHelper>();
> > +		helper->init(stages);
> > +
> > +		exposureModeHelpers_[exposureModeId] = helper;
> > +		availableExposureModes.push_back(exposureModeId);
> > +	}
> > +
> > +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> > +
> > +	return 0;
> > +}
> > +
> > +/**
> > + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> > + * \param[in] tuningData the YamlObject representing the tuning data
> > + *
> > + * This function parses tuning data to build the list of allowed values for the
> > + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> > + * the data in a specific format; the Agc algorithm's tuning data should contain
> > + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> > + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> > + * Each mode dict may contain either a "lower" or "upper" key or both, for
> > + * example:
> > + *
> > + * \code{.unparsed}
> > + * algorithms:
> > + *   - Agc:
> > + *       AeConstraintMode:
> > + *         ConstraintNormal:
> > + *           lower:
> > + *             qLo: 0.98
> > + *             qHi: 1.0
> > + *             yTarget: 0.5
> > + *         ConstraintHighlight:
> > + *           lower:
> > + *             qLo: 0.98
> > + *             qHi: 1.0
> > + *             yTarget: 0.5
> > + *           upper:
> > + *             qLo: 0.98
> > + *             qHi: 1.0
> > + *             yTarget: 0.8
> > + *
> > + * \endcode
> > + *
> > + * For the AeExposureMode control the data should contain a dictionary called
> > + * AeExposureMode containing per-mode setting dictionaries with the key being a
> > + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> > + * contain an array of shutter times with the key "shutter" and an array of gain
> > + * values with the key "gain", in this format:
> > + *
> > + * \code{.unparsed}
> > + * algorithms:
> > + *   - Agc:
> > + *       AeExposureMode:
> > + *         ExposureNormal:
> > + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> > + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> > + *         ExposureShort:
> > + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> > + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> > + *
> > + * \endcode
> > + *
> > + * @return 0 on success or a negative error code
> > + */
> > +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> > +{
> > +	int ret;
> > +
> > +	parseRelativeLuminanceTarget(tuningData);
> > +
> > +	ret = parseConstraintModes(tuningData);
> > +	if (ret)
> > +		return ret;
> > +
> > +	ret = parseExposureModes(tuningData);
> > +	if (ret)
> > +		return ret;
> > +
> > +	return 0;
> > +}
> > +
> > +/**
> > + * \brief configure the ExposureModeHelpers for this class
> > + * \param[in] minShutter Minimum shutter time to allow
> > + * \param[in] maxShutter Maximum shutter time to allow
> > + * \param[in] minGain Minimum gain to allow
> > + * \param[in] maxGain Maximum gain to allow
> > + *
> > + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> > + * ExposureModeHelper that has been created for this class.
> > + */
> > +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> > +						    utils::Duration maxShutter,
> > +						    double minGain,
> > +						    double maxGain)
> > +{
> > +	for (auto &[id, helper] : exposureModeHelpers_)
> > +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> > +}
> > +
> > +/**
> > + * \fn AgcMeanLuminance::constraintModes()
> > + * \brief Get the constraint modes that have been parsed from tuning data
> > + */
> > +
> > +/**
> > + * \fn AgcMeanLuminance::exposureModeHelpers()
> > + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> > + */
> > +
> > +/**
> > + * \fn AgcMeanLuminance::controls()
> > + * \brief Get the controls that have been generated after parsing tuning data
> > + */
> > +
> > +/**
> > + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> > + * \brief Estimate the luminance of an image, adjusted by a given gain
> > + * \param[in] gain The gain with which to adjust the luminance estimate
> > + *
> > + * This function estimates the average relative luminance of the frame that
> > + * would be output by the sensor if an additional \a gain was applied. It is a
> > + * pure virtual function because estimation of luminance is a hardware-specific
> > + * operation, which depends wholly on the format of the stats that are delivered
> > + * to libcamera from the ISP. Derived classes must implement an overriding
> > + * function that calculates the normalised mean luminance value across the
> > + * entire image.
> > + *
> > + * \return The normalised relative luminance of the image
> > + */
> > +
> > +/**
> > + * \brief Estimate the initial gain needed to achieve a relative luminance
> > + * target
> > + *
> > + * To account for non-linearity caused by saturation, the value needs to be
> > + * estimated in an iterative process, as multiplying by a gain will not increase
> > + * the relative luminance by the same factor if some image regions are saturated
> > + *
> > + * \return The calculated initial gain
> > + */
> > +double AgcMeanLuminance::estimateInitialGain()
> > +{
> > +	double yTarget = relativeLuminanceTarget_;
> > +	double yGain = 1.0;
> > +
> > +	for (unsigned int i = 0; i < 8; i++) {
> > +		double yValue = estimateLuminance(yGain);
> > +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> > +
> > +		yGain *= extra_gain;
> > +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> > +				<< ", Y target: " << yTarget
> > +				<< ", gives gain " << yGain;
> > +
> > +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> > +			break;
> > +	}
> > +
> > +	return yGain;
> > +}
> > +
> > +/**
> > + * \brief Clamp gain within the bounds of a defined constraint
> > + * \param[in] constraintModeIndex The index of the constraint to adhere to
> > + * \param[in] hist A histogram over which to calculate inter-quantile means
> > + * \param[in] gain The gain to clamp
> > + *
> > + * \return The gain clamped within the constraint bounds
> > + */
> > +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> > +					     const Histogram &hist,
> > +					     double gain)
> > +{
> > +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> > +	for (const AgcConstraint &constraint : constraints) {
> > +		double newGain = constraint.yTarget * hist.bins() /
> > +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> > +
> > +		if (constraint.bound == AgcConstraint::Bound::lower &&
> > +		    newGain > gain)
> > +			gain = newGain;
> > +
> > +		if (constraint.bound == AgcConstraint::Bound::upper &&
> > +		    newGain < gain)
> > +			gain = newGain;
> > +	}
> > +
> > +	return gain;
> > +}
> > +
> > +/**
> > + * \brief Apply a filter on the exposure value to limit the speed of changes
> > + * \param[in] exposureValue The target exposure from the AGC algorithm
> > + *
> > + * The speed of the filter is adaptive, and will produce the target quicker
> > + * during startup, or when the target exposure is within 20% of the most recent
> > + * filter output.
> > + *
> > + * \return The filtered exposure
> > + */
> > +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> > +{
> > +	double speed = 0.2;
> > +
> > +	/* Adapt instantly if we are in startup phase. */
> > +	if (frameCount_ < kNumStartupFrames)
> > +		speed = 1.0;
> > +
> > +	/*
> > +	 * If we are close to the desired result, go faster to avoid making
> > +	 * multiple micro-adjustments.
> > +	 * \todo Make this customisable?
> > +	 */
> > +	if (filteredExposure_ < 1.2 * exposureValue &&
> > +	    filteredExposure_ > 0.8 * exposureValue)
> > +		speed = sqrt(speed);
> > +
> > +	filteredExposure_ = speed * exposureValue +
> > +			    filteredExposure_ * (1.0 - speed);
> > +
> > +	return filteredExposure_;
> > +}
> > +
> > +/**
> > + * \brief Calculate the new exposure value
> > + * \param[in] constraintModeIndex The index of the current constraint mode
> > + * \param[in] exposureModeIndex The index of the current exposure mode
> > + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> > + *	      the calculated gain
> > + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> > + *	      statistics in use derive
> > + *
> > + * Calculate a new exposure value to try to obtain the target. The calculated
> > + * exposure value is filtered to prevent rapid changes from frame to frame, and
> > + * divided into shutter time, analogue and digital gain.
> > + *
> > + * \return Tuple of shutter time, analogue gain, and digital gain
> > + */
> > +std::tuple<utils::Duration, double, double>
> > +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> > +				 uint32_t exposureModeIndex,
> > +				 const Histogram &yHist,
> > +				 utils::Duration effectiveExposureValue)
> > +{
> > +	/*
> > +	 * The pipeline handler should validate that we have received an allowed
> > +	 * value for AeExposureMode.
> > +	 */
> > +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> > +		exposureModeHelpers_.at(exposureModeIndex);
> > +
> > +	double gain = estimateInitialGain();
> > +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> > +
> > +	/*
> > +	 * We don't check whether we're already close to the target, because
> > +	 * even if the effective exposure value is the same as the last frame's
> > +	 * we could have switched to an exposure mode that would require a new
> > +	 * pass through the splitExposure() function.
> > +	 */
> > +
> > +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> > +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> > +					   * exposureModeHelper->maxGain();
> > +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> > +
> > +	/*
> > +	 * We filter the exposure value to make sure changes are not too jarring
> > +	 * from frame to frame.
> > +	 */
> > +	newExposureValue = filterExposure(newExposureValue);
> > +
> > +	frameCount_++;
> > +	return exposureModeHelper->splitExposure(newExposureValue);
> > +}
> > +
> > +/**
> > + * \fn AgcMeanLuminance::resetFrameCount()
> > + * \brief Reset the frame counter
> > + *
> > + * This function resets the internal frame counter, which exists to help the
> > + * algorithm decide whether it should respond instantly or not. The expectation
> > + * is for derived classes to call this function before each camera start call,
> > + * either in configure() or queueRequest() if the frame number is zero.
> > + */
> > +
> > +}; /* namespace ipa */
> > +
> > +}; /* namespace libcamera */
> > diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> > new file mode 100644
> > index 00000000..e48dc498
> > --- /dev/null
> > +++ b/src/ipa/libipa/agc_mean_luminance.h
> > @@ -0,0 +1,91 @@
> > +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> > +/*
> > + * Copyright (C) 2024 Ideas on Board Oy
> > + *
> > + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> > + */
> > +
> > +#pragma once
> > +
> > +#include <tuple>
> > +#include <vector>
> > +
> > +#include <libcamera/controls.h>
> > +
> > +#include "libcamera/internal/yaml_parser.h"
> > +
> > +#include "exposure_mode_helper.h"
> > +#include "histogram.h"
> > +
> > +namespace libcamera {
> > +
> > +namespace ipa {
> > +
> > +class AgcMeanLuminance
> > +{
> > +public:
> > +	AgcMeanLuminance();
> > +	virtual ~AgcMeanLuminance() = default;
> > +
> > +	struct AgcConstraint {
> > +		enum class Bound {
> > +			lower = 0,
> > +			upper = 1
> > +		};
> > +		Bound bound;
> > +		double qLo;
> > +		double qHi;
> > +		double yTarget;
> > +	};
> > +
> > +	int parseTuningData(const YamlObject &tuningData);
> > +
> > +	void configureExposureModeHelpers(utils::Duration minShutter,
> > +					  utils::Duration maxShutter,
> > +					  double minGain,
> > +					  double maxGain);
> > +
> > +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> > +	{
> > +		return constraintModes_;
> > +	}
> > +
> > +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> > +	{
> > +		return exposureModeHelpers_;
> > +	}
> > +
> > +	ControlInfoMap::Map controls()
> > +	{
> > +		return controls_;
> > +	}
> > +
> > +	double estimateInitialGain();
> > +	double constraintClampGain(uint32_t constraintModeIndex,
> > +				   const Histogram &hist,
> > +				   double gain);
> > +	utils::Duration filterExposure(utils::Duration exposureValue);
> > +	std::tuple<utils::Duration, double, double>
> > +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> > +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> > +	void resetFrameCount() { frameCount_ = 0; }
> > +private:
> > +	virtual double estimateLuminance(const double gain) = 0;
> > +
> > +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> > +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> > +	int parseConstraintModes(const YamlObject &tuningData);
> > +	int parseExposureModes(const YamlObject &tuningData);
> > +
> > +	uint64_t frameCount_;
> > +	utils::Duration filteredExposure_;
> > +	double relativeLuminanceTarget_;
> > +
> > +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> > +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> > +	ControlInfoMap::Map controls_;
> > +};
> > +
> > +}; /* namespace ipa */
> > +
> > +}; /* namespace libcamera */
> > diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> > index 37fbd177..7ce885da 100644
> > --- a/src/ipa/libipa/meson.build
> > +++ b/src/ipa/libipa/meson.build
> > @@ -1,6 +1,7 @@
> >  # SPDX-License-Identifier: CC0-1.0
> >  
> >  libipa_headers = files([
> > +    'agc_mean_luminance.h',
> >      'algorithm.h',
> >      'camera_sensor_helper.h',
> >      'exposure_mode_helper.h',
> > @@ -10,6 +11,7 @@ libipa_headers = files([
> >  ])
> >  
> >  libipa_sources = files([
> > +    'agc_mean_luminance.cpp',
> >      'algorithm.cpp',
> >      'camera_sensor_helper.cpp',
> >      'exposure_mode_helper.cpp',
> > -- 
> > 2.34.1
> >

Hi Dan

On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> very large part; following the Rpi IPA's algorithm in spirit with a
> few tunable values in that IPA being hardcoded in the libipa ones.
> Add a new base class for MeanLuminanceAgc which implements the same
> algorithm and additionally parses yaml tuning files to inform an IPA
> module's Agc algorithm about valid constraint and exposure modes and
> their associated bounds.
>
> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> ---
> Changes in v2:
>
> 	- Renamed the class and files
> 	- Expanded the documentation
> 	- Added parseTuningData() so derived classes can call a single function
> 	  to cover all the parsing in ::init().
>
>  src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
>  src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
>  src/ipa/libipa/meson.build            |   2 +
>  3 files changed, 674 insertions(+)
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
>  create mode 100644 src/ipa/libipa/agc_mean_luminance.h
>
> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> new file mode 100644
> index 00000000..02e223cf
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> @@ -0,0 +1,581 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> + */
> +
> +#include "agc_mean_luminance.h"
> +
> +#include <cmath>
> +
> +#include <libcamera/base/log.h>
> +#include <libcamera/control_ids.h>
> +
> +#include "exposure_mode_helper.h"
> +
> +using namespace libcamera::controls;
> +
> +/**
> + * \file agc_mean_luminance.h
> + * \brief Base class implementing mean luminance AEGC
> + */
> +
> +namespace libcamera {
> +
> +using namespace std::literals::chrono_literals;
> +
> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> +
> +namespace ipa {
> +
> +/*
> + * Number of frames for which to run the algorithm at full speed, before slowing
> + * down to prevent large and jarring changes in exposure from frame to frame.
> + */
> +static constexpr uint32_t kNumStartupFrames = 10;
> +
> +/*
> + * Default relative luminance target
> + *
> + * This value should be chosen so that when the camera points at a grey target,
> + * the resulting image brightness looks "right". Custom values can be passed
> + * as the relativeLuminanceTarget value in sensor tuning files.
> + */
> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> +
> +/**
> + * \struct AgcMeanLuminance::AgcConstraint
> + * \brief The boundaries and target for an AeConstraintMode constraint
> + *
> + * This structure describes an AeConstraintMode constraint for the purposes of
> + * this algorithm. The algorithm will apply the constraints by calculating the
> + * Histogram's inter-quantile mean between the given quantiles and ensure that
> + * the resulting value is the right side of the given target (as defined by the
> + * boundary and luminance target).

I would appreciate a more generic overview of the what these
constraints are and what they express, instead of specifying how the
algorithm uses them (that's more appropriate for a comment in the code
if you like).

> + */
> +
> +/**
> + * \enum AgcMeanLuminance::AgcConstraint::Bound
> + * \brief Specify whether the constraint defines a lower or upper bound
> + * \var AgcMeanLuminance::AgcConstraint::lower
> + * \brief The constraint defines a lower bound
> + * \var AgcMeanLuminance::AgcConstraint::upper
> + * \brief The constraint defines an upper bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::bound
> + * \brief The type of constraint bound
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qLo
> + * \brief The lower quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::qHi
> + * \brief The upper quantile to use for the constraint
> + */
> +
> +/**
> + * \var AgcMeanLuminance::AgcConstraint::yTarget
> + * \brief The luminance target for the constraint
> + */
> +
> +/**
> + * \class AgcMeanLuminance
> + * \brief A mean-based auto-exposure algorithm
> + *
> + * This algorithm calculates a shutter time, analogue and digital gain such that
> + * the normalised mean luminance value of an image is driven towards a target,
> + * which itself is discovered from tuning data. The algorithm is a two-stage
> + * process.
> + *
> + * In the first stage, an initial gain value is derived by iteratively comparing
> + * the gain-adjusted mean luminance across an entire image against a target, and
> + * selecting a value which pushes it as closely as possible towards the target.
> + *
> + * In the second stage we calculate the gain required to drive the average of a
> + * section of a histogram to a target value, where the target and the boundaries
> + * of the section of the histogram used in the calculation are taken from the
> + * values defined for the currently configured AeConstraintMode within the
> + * tuning data. This class provides a helper function to parse those tuning data
> + * to discover the constraints, and so requires a specific format for those
> + * data which is described in \ref parseTuningData(). The gain from the first
> + * stage is then clamped to the gain from this stage.

This is a nice description of the internal working, and it has value
indeed.

However, the class API is not described, nor how IPA modules are
expected to use this class in their implementation. Do you think this
would have any value ?

> + *
> + * The final gain is used to adjust the effective exposure value of the image,
> + * and that new exposure value is divided into shutter time, analogue gain and
> + * digital gain according to the selected AeExposureMode. This class expects to

Why does it expects to ? doesn't it just do that ?

> + * use the \ref ExposureModeHelper class to assist in that division, and expects
> + * the data needed to initialise that class to be present in tuning data in a
> + * format described in \ref parseTuningData().
> + *
> + * In order to be able to derive an AGC implementation from this class, an IPA

Ah here you go, so IPA modules are expected to derive this class and
not use it by composition it seems. A little more details on this
might help IPA implementers to figure this out.

> + * needs to be able to do the following:
> + *
> + * 1. Provide a luminance estimation across an entire image.
> + * 2. Provide a luminance Histogram for the image to use in calculating
> + *    constraint compliance. The precision of the Histogram that is available
> + *    will determine the supportable precision of the constraints.
> + */
> +
> +AgcMeanLuminance::AgcMeanLuminance()
> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> +{
> +}
> +
> +/**
> + * \brief Parse the relative luminance target from the tuning data
> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> + */
> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> +{
> +	relativeLuminanceTarget_ =
> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> +}
> +
> +/**
> + * \brief Parse an AeConstraintMode constraint from tuning data
> + * \param[in] modeDict the YamlObject holding the constraint data
> + * \param[in] id The constraint ID from AeConstraintModeEnum
> + */
> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> +{
> +	for (const auto &[boundName, content] : modeDict.asDict()) {
> +		if (boundName != "upper" && boundName != "lower") {
> +			LOG(AgcMeanLuminance, Warning)
> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> +			continue;
> +		}
> +
> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> +		double qLo = content["qLo"].get<double>().value_or(0.98);
> +		double qHi = content["qHi"].get<double>().value_or(1.0);
> +		double yTarget =
> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> +
> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> +
> +		if (!constraintModes_.count(id))
> +			constraintModes_[id] = {};
> +
> +		if (idx)
> +			constraintModes_[id].push_back(constraint);
> +		else
> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> +	}
> +}
> +
> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableConstraintModes;
> +
> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> +	if (yamlConstraintModes.isDictionary()) {
> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> +			if (AeConstraintModeNameValueMap.find(modeName) ==
> +			    AeConstraintModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown constraint mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeDict.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid constraint mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			parseConstraint(modeDict,
> +					AeConstraintModeNameValueMap.at(modeName));
> +			availableConstraintModes.push_back(
> +				AeConstraintModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If the tuning data file contains no constraints then we use the
> +	 * default constraint that the various Agc algorithms were adhering to
> +	 * anyway before centralisation.

If you read this in 2 years, without knowing there was a
centralization and what "various algorithms" are, you want be able to
make a sense out of this.

What is needed here is to explain what the below values represent I
guess..

> +	 */
> +	if (constraintModes_.empty()) {
> +		AgcConstraint constraint = {
> +			AgcConstraint::Bound::lower,
> +			0.98,
> +			1.0,
> +			0.5
> +		};
> +
> +		constraintModes_[controls::ConstraintNormal].insert(
> +			constraintModes_[controls::ConstraintNormal].begin(),
> +			constraint);
> +		availableConstraintModes.push_back(
> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> +	}
> +
> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> +
> +	return 0;
> +}
> +
> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> +{
> +	std::vector<ControlValue> availableExposureModes;
> +
> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> +	if (yamlExposureModes.isDictionary()) {
> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> +			if (AeExposureModeNameValueMap.find(modeName) ==
> +			    AeExposureModeNameValueMap.end()) {
> +				LOG(AgcMeanLuminance, Warning)
> +					<< "Skipping unknown exposure mode '" << modeName << "'";
> +				continue;
> +			}
> +
> +			if (!modeValues.isDictionary()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Invalid exposure mode '" << modeName << "'";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<uint32_t> shutters =
> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> +			std::vector<double> gains =
> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> +
> +			if (shutters.size() != gains.size()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain array sizes unequal";
> +				return -EINVAL;
> +			}
> +
> +			if (shutters.empty()) {
> +				LOG(AgcMeanLuminance, Error)
> +					<< "Shutter and gain arrays are empty";
> +				return -EINVAL;
> +			}
> +
> +			std::vector<std::pair<utils::Duration, double>> stages;
> +			for (unsigned int i = 0; i < shutters.size(); i++) {
> +				stages.push_back({
> +					std::chrono::microseconds(shutters[i]),
> +					gains[i]
> +				});
> +			}
> +
> +			std::shared_ptr<ExposureModeHelper> helper =
> +				std::make_shared<ExposureModeHelper>();
> +			helper->init(stages);
> +
> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> +		}
> +	}
> +
> +	/*
> +	 * If we don't have any exposure modes in the tuning data we create an
> +	 * ExposureModeHelper using an empty vector of stages. This will result
> +	 * in the ExposureModeHelper simply driving the shutter as high as
> +	 * possible before touching gain.
> +	 */
> +	if (availableExposureModes.empty()) {
> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> +		std::vector<std::pair<utils::Duration, double>> stages = { };
> +
> +		std::shared_ptr<ExposureModeHelper> helper =
> +			std::make_shared<ExposureModeHelper>();
> +		helper->init(stages);
> +
> +		exposureModeHelpers_[exposureModeId] = helper;
> +		availableExposureModes.push_back(exposureModeId);
> +	}
> +
> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> + * \param[in] tuningData the YamlObject representing the tuning data
> + *
> + * This function parses tuning data to build the list of allowed values for the
> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> + * the data in a specific format; the Agc algorithm's tuning data should contain
> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> + * Each mode dict may contain either a "lower" or "upper" key or both, for
> + * example:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeConstraintMode:
> + *         ConstraintNormal:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *         ConstraintHighlight:
> + *           lower:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.5
> + *           upper:
> + *             qLo: 0.98
> + *             qHi: 1.0
> + *             yTarget: 0.8
> + *
> + * \endcode
> + *
> + * For the AeExposureMode control the data should contain a dictionary called
> + * AeExposureMode containing per-mode setting dictionaries with the key being a
> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> + * contain an array of shutter times with the key "shutter" and an array of gain
> + * values with the key "gain", in this format:
> + *
> + * \code{.unparsed}
> + * algorithms:
> + *   - Agc:
> + *       AeExposureMode:
> + *         ExposureNormal:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *         ExposureShort:
> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> + *
> + * \endcode
> + *
> + * @return 0 on success or a negative error code
> + */
> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> +{
> +	int ret;
> +
> +	parseRelativeLuminanceTarget(tuningData);
> +
> +	ret = parseConstraintModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	ret = parseExposureModes(tuningData);
> +	if (ret)
> +		return ret;
> +
> +	return 0;
> +}
> +
> +/**
> + * \brief configure the ExposureModeHelpers for this class
> + * \param[in] minShutter Minimum shutter time to allow
> + * \param[in] maxShutter Maximum shutter time to allow
> + * \param[in] minGain Minimum gain to allow
> + * \param[in] maxGain Maximum gain to allow
> + *
> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> + * ExposureModeHelper that has been created for this class.
> + */
> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> +						    utils::Duration maxShutter,
> +						    double minGain,
> +						    double maxGain)
> +{
> +	for (auto &[id, helper] : exposureModeHelpers_)
> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> +}
> +
> +/**
> + * \fn AgcMeanLuminance::constraintModes()
> + * \brief Get the constraint modes that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::exposureModeHelpers()
> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::controls()
> + * \brief Get the controls that have been generated after parsing tuning data
> + */
> +
> +/**
> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> + * \brief Estimate the luminance of an image, adjusted by a given gain
> + * \param[in] gain The gain with which to adjust the luminance estimate
> + *
> + * This function estimates the average relative luminance of the frame that
> + * would be output by the sensor if an additional \a gain was applied. It is a
> + * pure virtual function because estimation of luminance is a hardware-specific
> + * operation, which depends wholly on the format of the stats that are delivered
> + * to libcamera from the ISP. Derived classes must implement an overriding

or just "must override"

> + * function that calculates the normalised mean luminance value across the
> + * entire image.
> + *
> + * \return The normalised relative luminance of the image
> + */
> +
> +/**
> + * \brief Estimate the initial gain needed to achieve a relative luminance
> + * target
> + *
> + * To account for non-linearity caused by saturation, the value needs to be
> + * estimated in an iterative process, as multiplying by a gain will not increase
> + * the relative luminance by the same factor if some image regions are saturated

This seems more a comment to the above for loop than documentation

> + *
> + * \return The calculated initial gain
> + */
> +double AgcMeanLuminance::estimateInitialGain()
> +{
> +	double yTarget = relativeLuminanceTarget_;
> +	double yGain = 1.0;
> +
> +	for (unsigned int i = 0; i < 8; i++) {
> +		double yValue = estimateLuminance(yGain);
> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> +
> +		yGain *= extra_gain;
> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> +				<< ", Y target: " << yTarget
> +				<< ", gives gain " << yGain;
> +
> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> +			break;
> +	}
> +
> +	return yGain;
> +}
> +
> +/**
> + * \brief Clamp gain within the bounds of a defined constraint
> + * \param[in] constraintModeIndex The index of the constraint to adhere to
> + * \param[in] hist A histogram over which to calculate inter-quantile means
> + * \param[in] gain The gain to clamp
> + *
> + * \return The gain clamped within the constraint bounds
> + */
> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> +					     const Histogram &hist,
> +					     double gain)
> +{
> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> +	for (const AgcConstraint &constraint : constraints) {
> +		double newGain = constraint.yTarget * hist.bins() /
> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> +
> +		if (constraint.bound == AgcConstraint::Bound::lower &&
> +		    newGain > gain)
> +			gain = newGain;
> +
> +		if (constraint.bound == AgcConstraint::Bound::upper &&
> +		    newGain < gain)
> +			gain = newGain;
> +	}
> +
> +	return gain;
> +}
> +
> +/**
> + * \brief Apply a filter on the exposure value to limit the speed of changes
> + * \param[in] exposureValue The target exposure from the AGC algorithm
> + *
> + * The speed of the filter is adaptive, and will produce the target quicker
> + * during startup, or when the target exposure is within 20% of the most recent
> + * filter output.
> + *
> + * \return The filtered exposure
> + */
> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> +{
> +	double speed = 0.2;
> +
> +	/* Adapt instantly if we are in startup phase. */
> +	if (frameCount_ < kNumStartupFrames)
> +		speed = 1.0;
> +
> +	/*
> +	 * If we are close to the desired result, go faster to avoid making
> +	 * multiple micro-adjustments.
> +	 * \todo Make this customisable?
> +	 */
> +	if (filteredExposure_ < 1.2 * exposureValue &&
> +	    filteredExposure_ > 0.8 * exposureValue)
> +		speed = sqrt(speed);
> +
> +	filteredExposure_ = speed * exposureValue +
> +			    filteredExposure_ * (1.0 - speed);
> +
> +	return filteredExposure_;
> +}
> +
> +/**
> + * \brief Calculate the new exposure value
> + * \param[in] constraintModeIndex The index of the current constraint mode
> + * \param[in] exposureModeIndex The index of the current exposure mode
> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> + *	      the calculated gain
> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> + *	      statistics in use derive
> + *
> + * Calculate a new exposure value to try to obtain the target. The calculated
> + * exposure value is filtered to prevent rapid changes from frame to frame, and
> + * divided into shutter time, analogue and digital gain.
> + *
> + * \return Tuple of shutter time, analogue gain, and digital gain
> + */
> +std::tuple<utils::Duration, double, double>
> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> +				 uint32_t exposureModeIndex,
> +				 const Histogram &yHist,
> +				 utils::Duration effectiveExposureValue)
> +{
> +	/*
> +	 * The pipeline handler should validate that we have received an allowed
> +	 * value for AeExposureMode.

However, a check wouldn't hurt

> +	 */
> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> +		exposureModeHelpers_.at(exposureModeIndex);
> +
> +	double gain = estimateInitialGain();
> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> +
> +	/*
> +	 * We don't check whether we're already close to the target, because
> +	 * even if the effective exposure value is the same as the last frame's
> +	 * we could have switched to an exposure mode that would require a new
> +	 * pass through the splitExposure() function.
> +	 */
> +
> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> +					   * exposureModeHelper->maxGain();
> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> +
> +	/*
> +	 * We filter the exposure value to make sure changes are not too jarring
> +	 * from frame to frame.
> +	 */
> +	newExposureValue = filterExposure(newExposureValue);
> +
> +	frameCount_++;
> +	return exposureModeHelper->splitExposure(newExposureValue);

I wonder if this class shouldn't instead just calculate the exposure
value and let the IPA divide it up as it likes.. Just thinking out
loud and no real preference..

> +}
> +
> +/**
> + * \fn AgcMeanLuminance::resetFrameCount()
> + * \brief Reset the frame counter
> + *
> + * This function resets the internal frame counter, which exists to help the
> + * algorithm decide whether it should respond instantly or not. The expectation
> + * is for derived classes to call this function before each camera start call,
> + * either in configure() or queueRequest() if the frame number is zero.
> + */
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> new file mode 100644
> index 00000000..e48dc498
> --- /dev/null
> +++ b/src/ipa/libipa/agc_mean_luminance.h
> @@ -0,0 +1,91 @@
> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> +/*
> + * Copyright (C) 2024 Ideas on Board Oy
> + *
> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> + */
> +
> +#pragma once
> +
> +#include <tuple>
> +#include <vector>
> +
> +#include <libcamera/controls.h>
> +
> +#include "libcamera/internal/yaml_parser.h"
> +
> +#include "exposure_mode_helper.h"
> +#include "histogram.h"
> +
> +namespace libcamera {
> +
> +namespace ipa {
> +
> +class AgcMeanLuminance
> +{
> +public:
> +	AgcMeanLuminance();
> +	virtual ~AgcMeanLuminance() = default;
> +
> +	struct AgcConstraint {
> +		enum class Bound {
> +			lower = 0,
> +			upper = 1
> +		};
> +		Bound bound;
> +		double qLo;
> +		double qHi;
> +		double yTarget;
> +	};
> +
> +	int parseTuningData(const YamlObject &tuningData);
> +
> +	void configureExposureModeHelpers(utils::Duration minShutter,
> +					  utils::Duration maxShutter,
> +					  double minGain,
> +					  double maxGain);
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> +	{
> +		return constraintModes_;
> +	}
> +
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> +	{
> +		return exposureModeHelpers_;
> +	}
> +
> +	ControlInfoMap::Map controls()
> +	{
> +		return controls_;
> +	}
> +
> +	double estimateInitialGain();
> +	double constraintClampGain(uint32_t constraintModeIndex,
> +				   const Histogram &hist,
> +				   double gain);
> +	utils::Duration filterExposure(utils::Duration exposureValue);

Is filterExposure and clamps used outside of this class ?

> +	std::tuple<utils::Duration, double, double>
> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> +	void resetFrameCount() { frameCount_ = 0; }

Empty line

> +private:
> +	virtual double estimateLuminance(const double gain) = 0;
> +
> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> +	int parseConstraintModes(const YamlObject &tuningData);
> +	int parseExposureModes(const YamlObject &tuningData);
> +
> +	uint64_t frameCount_;
> +	utils::Duration filteredExposure_;
> +	double relativeLuminanceTarget_;
> +
> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> +	ControlInfoMap::Map controls_;
> +};
> +
> +}; /* namespace ipa */
> +
> +}; /* namespace libcamera */
> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> index 37fbd177..7ce885da 100644
> --- a/src/ipa/libipa/meson.build
> +++ b/src/ipa/libipa/meson.build
> @@ -1,6 +1,7 @@
>  # SPDX-License-Identifier: CC0-1.0
>
>  libipa_headers = files([
> +    'agc_mean_luminance.h',
>      'algorithm.h',
>      'camera_sensor_helper.h',
>      'exposure_mode_helper.h',
> @@ -10,6 +11,7 @@ libipa_headers = files([
>  ])
>
>  libipa_sources = files([
> +    'agc_mean_luminance.cpp',
>      'algorithm.cpp',
>      'camera_sensor_helper.cpp',
>      'exposure_mode_helper.cpp',
> --
> 2.34.1
>

Hi Stefan and Laurent

On 18/04/2024 08:48, Stefan Klug wrote:
> Hi Dan,
>
> thank you for the patch.
>
> On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
>> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
>> very large part; following the Rpi IPA's algorithm in spirit with a
>> few tunable values in that IPA being hardcoded in the libipa ones.
>> Add a new base class for MeanLuminanceAgc which implements the same
>> algorithm and additionally parses yaml tuning files to inform an IPA
>> module's Agc algorithm about valid constraint and exposure modes and
>> their associated bounds.
>>
>> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
>> ---
>> Changes in v2:
>>
>> 	- Renamed the class and files
>> 	- Expanded the documentation
>> 	- Added parseTuningData() so derived classes can call a single function
>> 	  to cover all the parsing in ::init().
>>
>>   src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
>>   src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
>>   src/ipa/libipa/meson.build            |   2 +
>>   3 files changed, 674 insertions(+)
>>   create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
>>   create mode 100644 src/ipa/libipa/agc_mean_luminance.h
>>
>> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
>> new file mode 100644
>> index 00000000..02e223cf
>> --- /dev/null
>> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
>> @@ -0,0 +1,581 @@
>> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
>> +/*
>> + * Copyright (C) 2024 Ideas on Board Oy
>> + *
>> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
>> + */
>> +
>> +#include "agc_mean_luminance.h"
>> +
>> +#include <cmath>
>> +
>> +#include <libcamera/base/log.h>
>> +#include <libcamera/control_ids.h>
>> +
>> +#include "exposure_mode_helper.h"
>> +
>> +using namespace libcamera::controls;
>> +
>> +/**
>> + * \file agc_mean_luminance.h
>> + * \brief Base class implementing mean luminance AEGC
>> + */
>> +
>> +namespace libcamera {
>> +
>> +using namespace std::literals::chrono_literals;
>> +
>> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
>> +
>> +namespace ipa {
>> +
>> +/*
>> + * Number of frames for which to run the algorithm at full speed, before slowing
>> + * down to prevent large and jarring changes in exposure from frame to frame.
>> + */
>> +static constexpr uint32_t kNumStartupFrames = 10;
>> +
>> +/*
>> + * Default relative luminance target
>> + *
>> + * This value should be chosen so that when the camera points at a grey target,
>> + * the resulting image brightness looks "right". Custom values can be passed
>> + * as the relativeLuminanceTarget value in sensor tuning files.
>> + */
>> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
>> +
>> +/**
>> + * \struct AgcMeanLuminance::AgcConstraint
>> + * \brief The boundaries and target for an AeConstraintMode constraint
>> + *
>> + * This structure describes an AeConstraintMode constraint for the purposes of
>> + * this algorithm. The algorithm will apply the constraints by calculating the
>> + * Histogram's inter-quantile mean between the given quantiles and ensure that
>> + * the resulting value is the right side of the given target (as defined by the
>> + * boundary and luminance target).
>> + */
>> +
>> +/**
>> + * \enum AgcMeanLuminance::AgcConstraint::Bound
>> + * \brief Specify whether the constraint defines a lower or upper bound
>> + * \var AgcMeanLuminance::AgcConstraint::lower
>> + * \brief The constraint defines a lower bound
>> + * \var AgcMeanLuminance::AgcConstraint::upper
>> + * \brief The constraint defines an upper bound
>> + */
>> +
>> +/**
>> + * \var AgcMeanLuminance::AgcConstraint::bound
>> + * \brief The type of constraint bound
>> + */
>> +
>> +/**
>> + * \var AgcMeanLuminance::AgcConstraint::qLo
>> + * \brief The lower quantile to use for the constraint
>> + */
>> +
>> +/**
>> + * \var AgcMeanLuminance::AgcConstraint::qHi
>> + * \brief The upper quantile to use for the constraint
>> + */
>> +
>> +/**
>> + * \var AgcMeanLuminance::AgcConstraint::yTarget
>> + * \brief The luminance target for the constraint
>> + */
>> +
>> +/**
>> + * \class AgcMeanLuminance
>> + * \brief A mean-based auto-exposure algorithm
>> + *
>> + * This algorithm calculates a shutter time, analogue and digital gain such that
>> + * the normalised mean luminance value of an image is driven towards a target,
>> + * which itself is discovered from tuning data. The algorithm is a two-stage
>> + * process.
>> + *
>> + * In the first stage, an initial gain value is derived by iteratively comparing
>> + * the gain-adjusted mean luminance across an entire image against a target, and
>> + * selecting a value which pushes it as closely as possible towards the target.
>> + *
>> + * In the second stage we calculate the gain required to drive the average of a
>> + * section of a histogram to a target value, where the target and the boundaries
>> + * of the section of the histogram used in the calculation are taken from the
>> + * values defined for the currently configured AeConstraintMode within the
>> + * tuning data. This class provides a helper function to parse those tuning data
>> + * to discover the constraints, and so requires a specific format for those
>> + * data which is described in \ref parseTuningData(). The gain from the first
>> + * stage is then clamped to the gain from this stage.
>> + *
>> + * The final gain is used to adjust the effective exposure value of the image,
>> + * and that new exposure value is divided into shutter time, analogue gain and
>> + * digital gain according to the selected AeExposureMode. This class expects to
>> + * use the \ref ExposureModeHelper class to assist in that division, and expects
>> + * the data needed to initialise that class to be present in tuning data in a
>> + * format described in \ref parseTuningData().
>> + *
>> + * In order to be able to derive an AGC implementation from this class, an IPA
>> + * needs to be able to do the following:
>> + *
>> + * 1. Provide a luminance estimation across an entire image.
>> + * 2. Provide a luminance Histogram for the image to use in calculating
>> + *    constraint compliance. The precision of the Histogram that is available
>> + *    will determine the supportable precision of the constraints.
>> + */
>> +
>> +AgcMeanLuminance::AgcMeanLuminance()
>> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
>> +{
>> +}
>> +
>> +/**
>> + * \brief Parse the relative luminance target from the tuning data
>> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
>> + */
>> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
>> +{
>> +	relativeLuminanceTarget_ =
>> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
>> +}
>> +
>> +/**
>> + * \brief Parse an AeConstraintMode constraint from tuning data
>> + * \param[in] modeDict the YamlObject holding the constraint data
>> + * \param[in] id The constraint ID from AeConstraintModeEnum
>> + */
>> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
>> +{
>> +	for (const auto &[boundName, content] : modeDict.asDict()) {
>> +		if (boundName != "upper" && boundName != "lower") {
>> +			LOG(AgcMeanLuminance, Warning)
>> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
>> +			continue;
>> +		}
>> +
>> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
>> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
>> +		double qLo = content["qLo"].get<double>().value_or(0.98);
>> +		double qHi = content["qHi"].get<double>().value_or(1.0);
>> +		double yTarget =
>> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
>> +
>> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
>> +
>> +		if (!constraintModes_.count(id))
>> +			constraintModes_[id] = {};
>> +
>> +		if (idx)
>> +			constraintModes_[id].push_back(constraint);
>> +		else
>> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
>> +	}
>> +}
>> +
>> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
>> +{
>> +	std::vector<ControlValue> availableConstraintModes;
>> +
>> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
>> +	if (yamlConstraintModes.isDictionary()) {
>> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
>> +			if (AeConstraintModeNameValueMap.find(modeName) ==
>> +			    AeConstraintModeNameValueMap.end()) {
>> +				LOG(AgcMeanLuminance, Warning)
>> +					<< "Skipping unknown constraint mode '" << modeName << "'";
>> +				continue;
>> +			}
>> +
>> +			if (!modeDict.isDictionary()) {
>> +				LOG(AgcMeanLuminance, Error)
>> +					<< "Invalid constraint mode '" << modeName << "'";
>> +				return -EINVAL;
>> +			}
>> +
>> +			parseConstraint(modeDict,
>> +					AeConstraintModeNameValueMap.at(modeName));
>> +			availableConstraintModes.push_back(
>> +				AeConstraintModeNameValueMap.at(modeName));
>> +		}
>> +	}
>> +
>> +	/*
>> +	 * If the tuning data file contains no constraints then we use the
>> +	 * default constraint that the various Agc algorithms were adhering to
>> +	 * anyway before centralisation.
>> +	 */
>> +	if (constraintModes_.empty()) {
>> +		AgcConstraint constraint = {
>> +			AgcConstraint::Bound::lower,
>> +			0.98,
>> +			1.0,
>> +			0.5
>> +		};
>> +
>> +		constraintModes_[controls::ConstraintNormal].insert(
>> +			constraintModes_[controls::ConstraintNormal].begin(),
>> +			constraint);
>> +		availableConstraintModes.push_back(
>> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
>> +	}
>> +
>> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
>> +
>> +	return 0;
>> +}
>> +
>> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
>> +{
>> +	std::vector<ControlValue> availableExposureModes;
>> +
>> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
>> +	if (yamlExposureModes.isDictionary()) {
>> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
>> +			if (AeExposureModeNameValueMap.find(modeName) ==
>> +			    AeExposureModeNameValueMap.end()) {
>> +				LOG(AgcMeanLuminance, Warning)
>> +					<< "Skipping unknown exposure mode '" << modeName << "'";
>> +				continue;
>> +			}
>> +
>> +			if (!modeValues.isDictionary()) {
>> +				LOG(AgcMeanLuminance, Error)
>> +					<< "Invalid exposure mode '" << modeName << "'";
>> +				return -EINVAL;
>> +			}
>> +
>> +			std::vector<uint32_t> shutters =
>> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
>> +			std::vector<double> gains =
>> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
>> +
>> +			if (shutters.size() != gains.size()) {
>> +				LOG(AgcMeanLuminance, Error)
>> +					<< "Shutter and gain array sizes unequal";
>> +				return -EINVAL;
>> +			}
>> +
>> +			if (shutters.empty()) {
>> +				LOG(AgcMeanLuminance, Error)
>> +					<< "Shutter and gain arrays are empty";
>> +				return -EINVAL;
>> +			}
>> +
>> +			std::vector<std::pair<utils::Duration, double>> stages;
>> +			for (unsigned int i = 0; i < shutters.size(); i++) {
>> +				stages.push_back({
>> +					std::chrono::microseconds(shutters[i]),
>> +					gains[i]
>> +				});
>> +			}
>> +
>> +			std::shared_ptr<ExposureModeHelper> helper =
>> +				std::make_shared<ExposureModeHelper>();
>> +			helper->init(stages);
>> +
>> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
>> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
>> +		}
>> +	}
>> +
>> +	/*
>> +	 * If we don't have any exposure modes in the tuning data we create an
>> +	 * ExposureModeHelper using an empty vector of stages. This will result
>> +	 * in the ExposureModeHelper simply driving the shutter as high as
>> +	 * possible before touching gain.
>> +	 */
>> +	if (availableExposureModes.empty()) {
>> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
>> +		std::vector<std::pair<utils::Duration, double>> stages = { };
>> +
>> +		std::shared_ptr<ExposureModeHelper> helper =
>> +			std::make_shared<ExposureModeHelper>();
>> +		helper->init(stages);
>> +
>> +		exposureModeHelpers_[exposureModeId] = helper;
>> +		availableExposureModes.push_back(exposureModeId);
>> +	}
>> +
>> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
>> +
>> +	return 0;
>> +}
>> +
>> +/**
>> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
>> + * \param[in] tuningData the YamlObject representing the tuning data
>> + *
>> + * This function parses tuning data to build the list of allowed values for the
>> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
>> + * the data in a specific format; the Agc algorithm's tuning data should contain
>> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
>> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
>> + * Each mode dict may contain either a "lower" or "upper" key or both, for
>> + * example:
>> + *
>> + * \code{.unparsed}
>> + * algorithms:
>> + *   - Agc:
>> + *       AeConstraintMode:
>> + *         ConstraintNormal:
>> + *           lower:
>> + *             qLo: 0.98
>> + *             qHi: 1.0
>> + *             yTarget: 0.5
>> + *         ConstraintHighlight:
>> + *           lower:
>> + *             qLo: 0.98
>> + *             qHi: 1.0
>> + *             yTarget: 0.5
>> + *           upper:
>> + *             qLo: 0.98
>> + *             qHi: 1.0
>> + *             yTarget: 0.8
>> + *
>> + * \endcode
>> + *
>> + * For the AeExposureMode control the data should contain a dictionary called
>> + * AeExposureMode containing per-mode setting dictionaries with the key being a
>> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
>> + * contain an array of shutter times with the key "shutter" and an array of gain
>> + * values with the key "gain", in this format:
>> + *
>> + * \code{.unparsed}
>> + * algorithms:
>> + *   - Agc:
>> + *       AeExposureMode:
>> + *         ExposureNormal:
>> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
>> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
>> + *         ExposureShort:
>> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
>> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
>> + *
>> + * \endcode
>> + *
>> + * @return 0 on success or a negative error code
>> + */
>> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
>> +{
>> +	int ret;
>> +
>> +	parseRelativeLuminanceTarget(tuningData);
>> +
>> +	ret = parseConstraintModes(tuningData);
>> +	if (ret)
>> +		return ret;
>> +
>> +	ret = parseExposureModes(tuningData);
>> +	if (ret)
>> +		return ret;
>> +
>> +	return 0;
>> +}
>> +
>> +/**
>> + * \brief configure the ExposureModeHelpers for this class
>> + * \param[in] minShutter Minimum shutter time to allow
>> + * \param[in] maxShutter Maximum shutter time to allow
>> + * \param[in] minGain Minimum gain to allow
>> + * \param[in] maxGain Maximum gain to allow
>> + *
>> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
>> + * ExposureModeHelper that has been created for this class.
>> + */
>> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
>> +						    utils::Duration maxShutter,
>> +						    double minGain,
>> +						    double maxGain)
>> +{
>> +	for (auto &[id, helper] : exposureModeHelpers_)
>> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
>> +}
>> +
>> +/**
>> + * \fn AgcMeanLuminance::constraintModes()
>> + * \brief Get the constraint modes that have been parsed from tuning data
>> + */
>> +
>> +/**
>> + * \fn AgcMeanLuminance::exposureModeHelpers()
>> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
>> + */
>> +
>> +/**
>> + * \fn AgcMeanLuminance::controls()
>> + * \brief Get the controls that have been generated after parsing tuning data
>> + */
>> +
>> +/**
>> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
>> + * \brief Estimate the luminance of an image, adjusted by a given gain
>> + * \param[in] gain The gain with which to adjust the luminance estimate
>> + *
>> + * This function estimates the average relative luminance of the frame that
>> + * would be output by the sensor if an additional \a gain was applied. It is a
>> + * pure virtual function because estimation of luminance is a hardware-specific
>> + * operation, which depends wholly on the format of the stats that are delivered
>> + * to libcamera from the ISP. Derived classes must implement an overriding
>> + * function that calculates the normalised mean luminance value across the
>> + * entire image.
>> + *
>> + * \return The normalised relative luminance of the image
>> + */
>> +
>> +/**
>> + * \brief Estimate the initial gain needed to achieve a relative luminance
>> + * target
>> + *
>> + * To account for non-linearity caused by saturation, the value needs to be
>> + * estimated in an iterative process, as multiplying by a gain will not increase
>> + * the relative luminance by the same factor if some image regions are saturated
>> + *
>> + * \return The calculated initial gain
>> + */
>> +double AgcMeanLuminance::estimateInitialGain()
>> +{
>> +	double yTarget = relativeLuminanceTarget_;
>> +	double yGain = 1.0;
>> +
>> +	for (unsigned int i = 0; i < 8; i++) {
>> +		double yValue = estimateLuminance(yGain);
>> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
>> +
>> +		yGain *= extra_gain;
>> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
>> +				<< ", Y target: " << yTarget
>> +				<< ", gives gain " << yGain;
>> +
>> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
>> +			break;
>> +	}
>> +
>> +	return yGain;
>> +}
>> +
>> +/**
>> + * \brief Clamp gain within the bounds of a defined constraint
>> + * \param[in] constraintModeIndex The index of the constraint to adhere to
>> + * \param[in] hist A histogram over which to calculate inter-quantile means
>> + * \param[in] gain The gain to clamp
>> + *
>> + * \return The gain clamped within the constraint bounds
>> + */
>> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
>> +					     const Histogram &hist,
>> +					     double gain)
>> +{
>> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
>> +	for (const AgcConstraint &constraint : constraints) {
>> +		double newGain = constraint.yTarget * hist.bins() /
>> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
>> +
>> +		if (constraint.bound == AgcConstraint::Bound::lower &&
>> +		    newGain > gain)
>> +			gain = newGain;
>> +
>> +		if (constraint.bound == AgcConstraint::Bound::upper &&
>> +		    newGain < gain)
>> +			gain = newGain;
>> +	}
>> +
>> +	return gain;
>> +}
>> +
>> +/**
>> + * \brief Apply a filter on the exposure value to limit the speed of changes
>> + * \param[in] exposureValue The target exposure from the AGC algorithm
>> + *
>> + * The speed of the filter is adaptive, and will produce the target quicker
>> + * during startup, or when the target exposure is within 20% of the most recent
>> + * filter output.
>> + *
>> + * \return The filtered exposure
>> + */
>> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
>> +{
>> +	double speed = 0.2;
>> +
>> +	/* Adapt instantly if we are in startup phase. */
>> +	if (frameCount_ < kNumStartupFrames)
>> +		speed = 1.0;
>> +
>> +	/*
>> +	 * If we are close to the desired result, go faster to avoid making
>> +	 * multiple micro-adjustments.
>> +	 * \todo Make this customisable?
>> +	 */
>> +	if (filteredExposure_ < 1.2 * exposureValue &&
>> +	    filteredExposure_ > 0.8 * exposureValue)
>> +		speed = sqrt(speed);
>> +
>> +	filteredExposure_ = speed * exposureValue +
>> +			    filteredExposure_ * (1.0 - speed);
>> +
>> +	return filteredExposure_;
>> +}
>> +
>> +/**
>> + * \brief Calculate the new exposure value
>> + * \param[in] constraintModeIndex The index of the current constraint mode
>> + * \param[in] exposureModeIndex The index of the current exposure mode
>> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
>> + *	      the calculated gain
> nit: no indentation
>
>> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
>> + *	      statistics in use derive
> nit: no indentation
>
>> + *
>> + * Calculate a new exposure value to try to obtain the target. The calculated
>> + * exposure value is filtered to prevent rapid changes from frame to frame, and
>> + * divided into shutter time, analogue and digital gain.
>> + *
>> + * \return Tuple of shutter time, analogue gain, and digital gain
>> + */
>> +std::tuple<utils::Duration, double, double>
>> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
>> +				 uint32_t exposureModeIndex,
>> +				 const Histogram &yHist,
>> +				 utils::Duration effectiveExposureValue)
>> +{
>> +	/*
>> +	 * The pipeline handler should validate that we have received an allowed
>> +	 * value for AeExposureMode.
>> +	 */
>> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
>> +		exposureModeHelpers_.at(exposureModeIndex);
>> +
>> +	double gain = estimateInitialGain();
>> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
>> +
>> +	/*
>> +	 * We don't check whether we're already close to the target, because
>> +	 * even if the effective exposure value is the same as the last frame's
>> +	 * we could have switched to an exposure mode that would require a new
>> +	 * pass through the splitExposure() function.
>> +	 */
>> +
>> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
>> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
>> +					   * exposureModeHelper->maxGain();
>> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
>> +
>> +	/*
>> +	 * We filter the exposure value to make sure changes are not too jarring
>> +	 * from frame to frame.
>> +	 */
>> +	newExposureValue = filterExposure(newExposureValue);
>> +
>> +	frameCount_++;
>> +	return exposureModeHelper->splitExposure(newExposureValue);
>> +}
>> +
>> +/**
>> + * \fn AgcMeanLuminance::resetFrameCount()
>> + * \brief Reset the frame counter
>> + *
>> + * This function resets the internal frame counter, which exists to help the
>> + * algorithm decide whether it should respond instantly or not. The expectation
>> + * is for derived classes to call this function before each camera start call,
>> + * either in configure() or queueRequest() if the frame number is zero.
>> + */
>> +
>> +}; /* namespace ipa */
>> +
>> +}; /* namespace libcamera */
>> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
>> new file mode 100644
>> index 00000000..e48dc498
>> --- /dev/null
>> +++ b/src/ipa/libipa/agc_mean_luminance.h
>> @@ -0,0 +1,91 @@
>> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
>> +/*
>> + * Copyright (C) 2024 Ideas on Board Oy
>> + *
>> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
>> + */
>> +
>> +#pragma once
>> +
>> +#include <tuple>
>> +#include <vector>
>> +
>> +#include <libcamera/controls.h>
>> +
>> +#include "libcamera/internal/yaml_parser.h"
>> +
>> +#include "exposure_mode_helper.h"
>> +#include "histogram.h"
>> +
>> +namespace libcamera {
>> +
>> +namespace ipa {
>> +
>> +class AgcMeanLuminance
>> +{
>> +public:
>> +	AgcMeanLuminance();
>> +	virtual ~AgcMeanLuminance() = default;
> There were a few small comments from Laurent that got lost
>   * destructor in cpp


The compiler says I can't follow this suggestion; since the estimateLuminance() function is virtual 
I apparently need a virtual destructor:


class libcamera::ipa::AgcMeanLuminance’ has virtual functions and accessible non-virtual destructor


Or am I doing something wrong?

>   * code sytel in enum


Isn't this addressed by making them lowercase?

>   * missing line
>
> Aside from that, I think we should merge it in.
>
> Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
>
> Cheers,
> Stefan
>
>> +
>> +	struct AgcConstraint {
>> +		enum class Bound {
>> +			lower = 0,
>> +			upper = 1
>> +		};
>> +		Bound bound;
>> +		double qLo;
>> +		double qHi;
>> +		double yTarget;
>> +	};
>> +
>> +	int parseTuningData(const YamlObject &tuningData);
>> +
>> +	void configureExposureModeHelpers(utils::Duration minShutter,
>> +					  utils::Duration maxShutter,
>> +					  double minGain,
>> +					  double maxGain);
>> +
>> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
>> +	{
>> +		return constraintModes_;
>> +	}
>> +
>> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
>> +	{
>> +		return exposureModeHelpers_;
>> +	}
>> +
>> +	ControlInfoMap::Map controls()
>> +	{
>> +		return controls_;
>> +	}
>> +
>> +	double estimateInitialGain();
>> +	double constraintClampGain(uint32_t constraintModeIndex,
>> +				   const Histogram &hist,
>> +				   double gain);
>> +	utils::Duration filterExposure(utils::Duration exposureValue);
>> +	std::tuple<utils::Duration, double, double>
>> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
>> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
>> +	void resetFrameCount() { frameCount_ = 0; }
>> +private:
>> +	virtual double estimateLuminance(const double gain) = 0;
>> +
>> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
>> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
>> +	int parseConstraintModes(const YamlObject &tuningData);
>> +	int parseExposureModes(const YamlObject &tuningData);
>> +
>> +	uint64_t frameCount_;
>> +	utils::Duration filteredExposure_;
>> +	double relativeLuminanceTarget_;
>> +
>> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
>> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
>> +	ControlInfoMap::Map controls_;
>> +};
>> +
>> +}; /* namespace ipa */
>> +
>> +}; /* namespace libcamera */
>> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
>> index 37fbd177..7ce885da 100644
>> --- a/src/ipa/libipa/meson.build
>> +++ b/src/ipa/libipa/meson.build
>> @@ -1,6 +1,7 @@
>>   # SPDX-License-Identifier: CC0-1.0
>>   
>>   libipa_headers = files([
>> +    'agc_mean_luminance.h',
>>       'algorithm.h',
>>       'camera_sensor_helper.h',
>>       'exposure_mode_helper.h',
>> @@ -10,6 +11,7 @@ libipa_headers = files([
>>   ])
>>   
>>   libipa_sources = files([
>> +    'agc_mean_luminance.cpp',
>>       'algorithm.cpp',
>>       'camera_sensor_helper.cpp',
>>       'exposure_mode_helper.cpp',
>> -- 
>> 2.34.1
>>

Hi Dan,

On Wed, Apr 24, 2024 at 09:51:58AM +0100, Daniel Scally wrote:
> On 18/04/2024 08:48, Stefan Klug wrote:
> > On Wed, Apr 17, 2024 at 02:15:32PM +0100, Daniel Scally wrote:
> >> The Agc algorithms for the RkIsp1 and IPU3 IPAs do the same thing in
> >> very large part; following the Rpi IPA's algorithm in spirit with a
> >> few tunable values in that IPA being hardcoded in the libipa ones.
> >> Add a new base class for MeanLuminanceAgc which implements the same
> >> algorithm and additionally parses yaml tuning files to inform an IPA
> >> module's Agc algorithm about valid constraint and exposure modes and
> >> their associated bounds.
> >>
> >> Signed-off-by: Daniel Scally <dan.scally@ideasonboard.com>
> >> ---
> >> Changes in v2:
> >>
> >> 	- Renamed the class and files
> >> 	- Expanded the documentation
> >> 	- Added parseTuningData() so derived classes can call a single function
> >> 	  to cover all the parsing in ::init().
> >>
> >>   src/ipa/libipa/agc_mean_luminance.cpp | 581 ++++++++++++++++++++++++++
> >>   src/ipa/libipa/agc_mean_luminance.h   |  91 ++++
> >>   src/ipa/libipa/meson.build            |   2 +
> >>   3 files changed, 674 insertions(+)
> >>   create mode 100644 src/ipa/libipa/agc_mean_luminance.cpp
> >>   create mode 100644 src/ipa/libipa/agc_mean_luminance.h
> >>
> >> diff --git a/src/ipa/libipa/agc_mean_luminance.cpp b/src/ipa/libipa/agc_mean_luminance.cpp
> >> new file mode 100644
> >> index 00000000..02e223cf
> >> --- /dev/null
> >> +++ b/src/ipa/libipa/agc_mean_luminance.cpp
> >> @@ -0,0 +1,581 @@
> >> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> >> +/*
> >> + * Copyright (C) 2024 Ideas on Board Oy
> >> + *
> >> + * agc_mean_luminance.cpp - Base class for mean luminance AGC algorithms
> >> + */
> >> +
> >> +#include "agc_mean_luminance.h"
> >> +
> >> +#include <cmath>
> >> +
> >> +#include <libcamera/base/log.h>
> >> +#include <libcamera/control_ids.h>
> >> +
> >> +#include "exposure_mode_helper.h"
> >> +
> >> +using namespace libcamera::controls;
> >> +
> >> +/**
> >> + * \file agc_mean_luminance.h
> >> + * \brief Base class implementing mean luminance AEGC
> >> + */
> >> +
> >> +namespace libcamera {
> >> +
> >> +using namespace std::literals::chrono_literals;
> >> +
> >> +LOG_DEFINE_CATEGORY(AgcMeanLuminance)
> >> +
> >> +namespace ipa {
> >> +
> >> +/*
> >> + * Number of frames for which to run the algorithm at full speed, before slowing
> >> + * down to prevent large and jarring changes in exposure from frame to frame.
> >> + */
> >> +static constexpr uint32_t kNumStartupFrames = 10;
> >> +
> >> +/*
> >> + * Default relative luminance target
> >> + *
> >> + * This value should be chosen so that when the camera points at a grey target,
> >> + * the resulting image brightness looks "right". Custom values can be passed
> >> + * as the relativeLuminanceTarget value in sensor tuning files.
> >> + */
> >> +static constexpr double kDefaultRelativeLuminanceTarget = 0.16;
> >> +
> >> +/**
> >> + * \struct AgcMeanLuminance::AgcConstraint
> >> + * \brief The boundaries and target for an AeConstraintMode constraint
> >> + *
> >> + * This structure describes an AeConstraintMode constraint for the purposes of
> >> + * this algorithm. The algorithm will apply the constraints by calculating the
> >> + * Histogram's inter-quantile mean between the given quantiles and ensure that
> >> + * the resulting value is the right side of the given target (as defined by the
> >> + * boundary and luminance target).
> >> + */
> >> +
> >> +/**
> >> + * \enum AgcMeanLuminance::AgcConstraint::Bound
> >> + * \brief Specify whether the constraint defines a lower or upper bound
> >> + * \var AgcMeanLuminance::AgcConstraint::lower
> >> + * \brief The constraint defines a lower bound
> >> + * \var AgcMeanLuminance::AgcConstraint::upper
> >> + * \brief The constraint defines an upper bound
> >> + */
> >> +
> >> +/**
> >> + * \var AgcMeanLuminance::AgcConstraint::bound
> >> + * \brief The type of constraint bound
> >> + */
> >> +
> >> +/**
> >> + * \var AgcMeanLuminance::AgcConstraint::qLo
> >> + * \brief The lower quantile to use for the constraint
> >> + */
> >> +
> >> +/**
> >> + * \var AgcMeanLuminance::AgcConstraint::qHi
> >> + * \brief The upper quantile to use for the constraint
> >> + */
> >> +
> >> +/**
> >> + * \var AgcMeanLuminance::AgcConstraint::yTarget
> >> + * \brief The luminance target for the constraint
> >> + */
> >> +
> >> +/**
> >> + * \class AgcMeanLuminance
> >> + * \brief A mean-based auto-exposure algorithm
> >> + *
> >> + * This algorithm calculates a shutter time, analogue and digital gain such that
> >> + * the normalised mean luminance value of an image is driven towards a target,
> >> + * which itself is discovered from tuning data. The algorithm is a two-stage
> >> + * process.
> >> + *
> >> + * In the first stage, an initial gain value is derived by iteratively comparing
> >> + * the gain-adjusted mean luminance across an entire image against a target, and
> >> + * selecting a value which pushes it as closely as possible towards the target.
> >> + *
> >> + * In the second stage we calculate the gain required to drive the average of a
> >> + * section of a histogram to a target value, where the target and the boundaries
> >> + * of the section of the histogram used in the calculation are taken from the
> >> + * values defined for the currently configured AeConstraintMode within the
> >> + * tuning data. This class provides a helper function to parse those tuning data
> >> + * to discover the constraints, and so requires a specific format for those
> >> + * data which is described in \ref parseTuningData(). The gain from the first
> >> + * stage is then clamped to the gain from this stage.
> >> + *
> >> + * The final gain is used to adjust the effective exposure value of the image,
> >> + * and that new exposure value is divided into shutter time, analogue gain and
> >> + * digital gain according to the selected AeExposureMode. This class expects to
> >> + * use the \ref ExposureModeHelper class to assist in that division, and expects
> >> + * the data needed to initialise that class to be present in tuning data in a
> >> + * format described in \ref parseTuningData().
> >> + *
> >> + * In order to be able to derive an AGC implementation from this class, an IPA
> >> + * needs to be able to do the following:
> >> + *
> >> + * 1. Provide a luminance estimation across an entire image.
> >> + * 2. Provide a luminance Histogram for the image to use in calculating
> >> + *    constraint compliance. The precision of the Histogram that is available
> >> + *    will determine the supportable precision of the constraints.
> >> + */
> >> +
> >> +AgcMeanLuminance::AgcMeanLuminance()
> >> +	: frameCount_(0), filteredExposure_(0s), relativeLuminanceTarget_(0)
> >> +{
> >> +}
> >> +
> >> +/**
> >> + * \brief Parse the relative luminance target from the tuning data
> >> + * \param[in] tuningData The YamlObject holding the algorithm's tuning data
> >> + */
> >> +void AgcMeanLuminance::parseRelativeLuminanceTarget(const YamlObject &tuningData)
> >> +{
> >> +	relativeLuminanceTarget_ =
> >> +		tuningData["relativeLuminanceTarget"].get<double>(kDefaultRelativeLuminanceTarget);
> >> +}
> >> +
> >> +/**
> >> + * \brief Parse an AeConstraintMode constraint from tuning data
> >> + * \param[in] modeDict the YamlObject holding the constraint data
> >> + * \param[in] id The constraint ID from AeConstraintModeEnum
> >> + */
> >> +void AgcMeanLuminance::parseConstraint(const YamlObject &modeDict, int32_t id)
> >> +{
> >> +	for (const auto &[boundName, content] : modeDict.asDict()) {
> >> +		if (boundName != "upper" && boundName != "lower") {
> >> +			LOG(AgcMeanLuminance, Warning)
> >> +				<< "Ignoring unknown constraint bound '" << boundName << "'";
> >> +			continue;
> >> +		}
> >> +
> >> +		unsigned int idx = static_cast<unsigned int>(boundName == "upper");
> >> +		AgcConstraint::Bound bound = static_cast<AgcConstraint::Bound>(idx);
> >> +		double qLo = content["qLo"].get<double>().value_or(0.98);
> >> +		double qHi = content["qHi"].get<double>().value_or(1.0);
> >> +		double yTarget =
> >> +			content["yTarget"].getList<double>().value_or(std::vector<double>{ 0.5 }).at(0);
> >> +
> >> +		AgcConstraint constraint = { bound, qLo, qHi, yTarget };
> >> +
> >> +		if (!constraintModes_.count(id))
> >> +			constraintModes_[id] = {};
> >> +
> >> +		if (idx)
> >> +			constraintModes_[id].push_back(constraint);
> >> +		else
> >> +			constraintModes_[id].insert(constraintModes_[id].begin(), constraint);
> >> +	}
> >> +}
> >> +
> >> +int AgcMeanLuminance::parseConstraintModes(const YamlObject &tuningData)
> >> +{
> >> +	std::vector<ControlValue> availableConstraintModes;
> >> +
> >> +	const YamlObject &yamlConstraintModes = tuningData[controls::AeConstraintMode.name()];
> >> +	if (yamlConstraintModes.isDictionary()) {
> >> +		for (const auto &[modeName, modeDict] : yamlConstraintModes.asDict()) {
> >> +			if (AeConstraintModeNameValueMap.find(modeName) ==
> >> +			    AeConstraintModeNameValueMap.end()) {
> >> +				LOG(AgcMeanLuminance, Warning)
> >> +					<< "Skipping unknown constraint mode '" << modeName << "'";
> >> +				continue;
> >> +			}
> >> +
> >> +			if (!modeDict.isDictionary()) {
> >> +				LOG(AgcMeanLuminance, Error)
> >> +					<< "Invalid constraint mode '" << modeName << "'";
> >> +				return -EINVAL;
> >> +			}
> >> +
> >> +			parseConstraint(modeDict,
> >> +					AeConstraintModeNameValueMap.at(modeName));
> >> +			availableConstraintModes.push_back(
> >> +				AeConstraintModeNameValueMap.at(modeName));
> >> +		}
> >> +	}
> >> +
> >> +	/*
> >> +	 * If the tuning data file contains no constraints then we use the
> >> +	 * default constraint that the various Agc algorithms were adhering to
> >> +	 * anyway before centralisation.
> >> +	 */
> >> +	if (constraintModes_.empty()) {
> >> +		AgcConstraint constraint = {
> >> +			AgcConstraint::Bound::lower,
> >> +			0.98,
> >> +			1.0,
> >> +			0.5
> >> +		};
> >> +
> >> +		constraintModes_[controls::ConstraintNormal].insert(
> >> +			constraintModes_[controls::ConstraintNormal].begin(),
> >> +			constraint);
> >> +		availableConstraintModes.push_back(
> >> +			AeConstraintModeNameValueMap.at("ConstraintNormal"));
> >> +	}
> >> +
> >> +	controls_[&controls::AeConstraintMode] = ControlInfo(availableConstraintModes);
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +int AgcMeanLuminance::parseExposureModes(const YamlObject &tuningData)
> >> +{
> >> +	std::vector<ControlValue> availableExposureModes;
> >> +
> >> +	const YamlObject &yamlExposureModes = tuningData[controls::AeExposureMode.name()];
> >> +	if (yamlExposureModes.isDictionary()) {
> >> +		for (const auto &[modeName, modeValues] : yamlExposureModes.asDict()) {
> >> +			if (AeExposureModeNameValueMap.find(modeName) ==
> >> +			    AeExposureModeNameValueMap.end()) {
> >> +				LOG(AgcMeanLuminance, Warning)
> >> +					<< "Skipping unknown exposure mode '" << modeName << "'";
> >> +				continue;
> >> +			}
> >> +
> >> +			if (!modeValues.isDictionary()) {
> >> +				LOG(AgcMeanLuminance, Error)
> >> +					<< "Invalid exposure mode '" << modeName << "'";
> >> +				return -EINVAL;
> >> +			}
> >> +
> >> +			std::vector<uint32_t> shutters =
> >> +				modeValues["shutter"].getList<uint32_t>().value_or(std::vector<uint32_t>{});
> >> +			std::vector<double> gains =
> >> +				modeValues["gain"].getList<double>().value_or(std::vector<double>{});
> >> +
> >> +			if (shutters.size() != gains.size()) {
> >> +				LOG(AgcMeanLuminance, Error)
> >> +					<< "Shutter and gain array sizes unequal";
> >> +				return -EINVAL;
> >> +			}
> >> +
> >> +			if (shutters.empty()) {
> >> +				LOG(AgcMeanLuminance, Error)
> >> +					<< "Shutter and gain arrays are empty";
> >> +				return -EINVAL;
> >> +			}
> >> +
> >> +			std::vector<std::pair<utils::Duration, double>> stages;
> >> +			for (unsigned int i = 0; i < shutters.size(); i++) {
> >> +				stages.push_back({
> >> +					std::chrono::microseconds(shutters[i]),
> >> +					gains[i]
> >> +				});
> >> +			}
> >> +
> >> +			std::shared_ptr<ExposureModeHelper> helper =
> >> +				std::make_shared<ExposureModeHelper>();
> >> +			helper->init(stages);
> >> +
> >> +			exposureModeHelpers_[AeExposureModeNameValueMap.at(modeName)] = helper;
> >> +			availableExposureModes.push_back(AeExposureModeNameValueMap.at(modeName));
> >> +		}
> >> +	}
> >> +
> >> +	/*
> >> +	 * If we don't have any exposure modes in the tuning data we create an
> >> +	 * ExposureModeHelper using an empty vector of stages. This will result
> >> +	 * in the ExposureModeHelper simply driving the shutter as high as
> >> +	 * possible before touching gain.
> >> +	 */
> >> +	if (availableExposureModes.empty()) {
> >> +		int32_t exposureModeId = AeExposureModeNameValueMap.at("ExposureNormal");
> >> +		std::vector<std::pair<utils::Duration, double>> stages = { };
> >> +
> >> +		std::shared_ptr<ExposureModeHelper> helper =
> >> +			std::make_shared<ExposureModeHelper>();
> >> +		helper->init(stages);
> >> +
> >> +		exposureModeHelpers_[exposureModeId] = helper;
> >> +		availableExposureModes.push_back(exposureModeId);
> >> +	}
> >> +
> >> +	controls_[&controls::AeExposureMode] = ControlInfo(availableExposureModes);
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +/**
> >> + * \brief Parse tuning data for AeConstraintMode and AeExposureMode controls
> >> + * \param[in] tuningData the YamlObject representing the tuning data
> >> + *
> >> + * This function parses tuning data to build the list of allowed values for the
> >> + * AeConstraintMode and AeExposureMode controls. Those tuning data must provide
> >> + * the data in a specific format; the Agc algorithm's tuning data should contain
> >> + * a dictionary called AeConstraintMode containing per-mode setting dictionaries
> >> + * with the key being a value from \ref controls::AeConstraintModeNameValueMap.
> >> + * Each mode dict may contain either a "lower" or "upper" key or both, for
> >> + * example:
> >> + *
> >> + * \code{.unparsed}
> >> + * algorithms:
> >> + *   - Agc:
> >> + *       AeConstraintMode:
> >> + *         ConstraintNormal:
> >> + *           lower:
> >> + *             qLo: 0.98
> >> + *             qHi: 1.0
> >> + *             yTarget: 0.5
> >> + *         ConstraintHighlight:
> >> + *           lower:
> >> + *             qLo: 0.98
> >> + *             qHi: 1.0
> >> + *             yTarget: 0.5
> >> + *           upper:
> >> + *             qLo: 0.98
> >> + *             qHi: 1.0
> >> + *             yTarget: 0.8
> >> + *
> >> + * \endcode
> >> + *
> >> + * For the AeExposureMode control the data should contain a dictionary called
> >> + * AeExposureMode containing per-mode setting dictionaries with the key being a
> >> + * value from \ref controls::AeExposureModeNameValueMap. Each mode dict should
> >> + * contain an array of shutter times with the key "shutter" and an array of gain
> >> + * values with the key "gain", in this format:
> >> + *
> >> + * \code{.unparsed}
> >> + * algorithms:
> >> + *   - Agc:
> >> + *       AeExposureMode:
> >> + *         ExposureNormal:
> >> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> >> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> >> + *         ExposureShort:
> >> + *           shutter: [ 100, 10000, 30000, 60000, 120000 ]
> >> + *           gain: [ 2.0, 4.0, 6.0, 8.0, 10.0 ]
> >> + *
> >> + * \endcode
> >> + *
> >> + * @return 0 on success or a negative error code
> >> + */
> >> +int AgcMeanLuminance::parseTuningData(const YamlObject &tuningData)
> >> +{
> >> +	int ret;
> >> +
> >> +	parseRelativeLuminanceTarget(tuningData);
> >> +
> >> +	ret = parseConstraintModes(tuningData);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	ret = parseExposureModes(tuningData);
> >> +	if (ret)
> >> +		return ret;
> >> +
> >> +	return 0;
> >> +}
> >> +
> >> +/**
> >> + * \brief configure the ExposureModeHelpers for this class
> >> + * \param[in] minShutter Minimum shutter time to allow
> >> + * \param[in] maxShutter Maximum shutter time to allow
> >> + * \param[in] minGain Minimum gain to allow
> >> + * \param[in] maxGain Maximum gain to allow
> >> + *
> >> + * This function calls \ref ExposureModeHelper::setShutterGainLimits() for each
> >> + * ExposureModeHelper that has been created for this class.
> >> + */
> >> +void AgcMeanLuminance::configureExposureModeHelpers(utils::Duration minShutter,
> >> +						    utils::Duration maxShutter,
> >> +						    double minGain,
> >> +						    double maxGain)
> >> +{
> >> +	for (auto &[id, helper] : exposureModeHelpers_)
> >> +		helper->setShutterGainLimits(minShutter, maxShutter, minGain, maxGain);
> >> +}
> >> +
> >> +/**
> >> + * \fn AgcMeanLuminance::constraintModes()
> >> + * \brief Get the constraint modes that have been parsed from tuning data
> >> + */
> >> +
> >> +/**
> >> + * \fn AgcMeanLuminance::exposureModeHelpers()
> >> + * \brief Get the ExposureModeHelpers that have been parsed from tuning data
> >> + */
> >> +
> >> +/**
> >> + * \fn AgcMeanLuminance::controls()
> >> + * \brief Get the controls that have been generated after parsing tuning data
> >> + */
> >> +
> >> +/**
> >> + * \fn AgcMeanLuminance::estimateLuminance(const double gain)
> >> + * \brief Estimate the luminance of an image, adjusted by a given gain
> >> + * \param[in] gain The gain with which to adjust the luminance estimate
> >> + *
> >> + * This function estimates the average relative luminance of the frame that
> >> + * would be output by the sensor if an additional \a gain was applied. It is a
> >> + * pure virtual function because estimation of luminance is a hardware-specific
> >> + * operation, which depends wholly on the format of the stats that are delivered
> >> + * to libcamera from the ISP. Derived classes must implement an overriding
> >> + * function that calculates the normalised mean luminance value across the
> >> + * entire image.
> >> + *
> >> + * \return The normalised relative luminance of the image
> >> + */
> >> +
> >> +/**
> >> + * \brief Estimate the initial gain needed to achieve a relative luminance
> >> + * target
> >> + *
> >> + * To account for non-linearity caused by saturation, the value needs to be
> >> + * estimated in an iterative process, as multiplying by a gain will not increase
> >> + * the relative luminance by the same factor if some image regions are saturated
> >> + *
> >> + * \return The calculated initial gain
> >> + */
> >> +double AgcMeanLuminance::estimateInitialGain()
> >> +{
> >> +	double yTarget = relativeLuminanceTarget_;
> >> +	double yGain = 1.0;
> >> +
> >> +	for (unsigned int i = 0; i < 8; i++) {
> >> +		double yValue = estimateLuminance(yGain);
> >> +		double extra_gain = std::min(10.0, yTarget / (yValue + .001));
> >> +
> >> +		yGain *= extra_gain;
> >> +		LOG(AgcMeanLuminance, Debug) << "Y value: " << yValue
> >> +				<< ", Y target: " << yTarget
> >> +				<< ", gives gain " << yGain;
> >> +
> >> +		if (utils::abs_diff(extra_gain, 1.0) < 0.01)
> >> +			break;
> >> +	}
> >> +
> >> +	return yGain;
> >> +}
> >> +
> >> +/**
> >> + * \brief Clamp gain within the bounds of a defined constraint
> >> + * \param[in] constraintModeIndex The index of the constraint to adhere to
> >> + * \param[in] hist A histogram over which to calculate inter-quantile means
> >> + * \param[in] gain The gain to clamp
> >> + *
> >> + * \return The gain clamped within the constraint bounds
> >> + */
> >> +double AgcMeanLuminance::constraintClampGain(uint32_t constraintModeIndex,
> >> +					     const Histogram &hist,
> >> +					     double gain)
> >> +{
> >> +	std::vector<AgcConstraint> &constraints = constraintModes_[constraintModeIndex];
> >> +	for (const AgcConstraint &constraint : constraints) {
> >> +		double newGain = constraint.yTarget * hist.bins() /
> >> +				 hist.interQuantileMean(constraint.qLo, constraint.qHi);
> >> +
> >> +		if (constraint.bound == AgcConstraint::Bound::lower &&
> >> +		    newGain > gain)
> >> +			gain = newGain;
> >> +
> >> +		if (constraint.bound == AgcConstraint::Bound::upper &&
> >> +		    newGain < gain)
> >> +			gain = newGain;
> >> +	}
> >> +
> >> +	return gain;
> >> +}
> >> +
> >> +/**
> >> + * \brief Apply a filter on the exposure value to limit the speed of changes
> >> + * \param[in] exposureValue The target exposure from the AGC algorithm
> >> + *
> >> + * The speed of the filter is adaptive, and will produce the target quicker
> >> + * during startup, or when the target exposure is within 20% of the most recent
> >> + * filter output.
> >> + *
> >> + * \return The filtered exposure
> >> + */
> >> +utils::Duration AgcMeanLuminance::filterExposure(utils::Duration exposureValue)
> >> +{
> >> +	double speed = 0.2;
> >> +
> >> +	/* Adapt instantly if we are in startup phase. */
> >> +	if (frameCount_ < kNumStartupFrames)
> >> +		speed = 1.0;
> >> +
> >> +	/*
> >> +	 * If we are close to the desired result, go faster to avoid making
> >> +	 * multiple micro-adjustments.
> >> +	 * \todo Make this customisable?
> >> +	 */
> >> +	if (filteredExposure_ < 1.2 * exposureValue &&
> >> +	    filteredExposure_ > 0.8 * exposureValue)
> >> +		speed = sqrt(speed);
> >> +
> >> +	filteredExposure_ = speed * exposureValue +
> >> +			    filteredExposure_ * (1.0 - speed);
> >> +
> >> +	return filteredExposure_;
> >> +}
> >> +
> >> +/**
> >> + * \brief Calculate the new exposure value
> >> + * \param[in] constraintModeIndex The index of the current constraint mode
> >> + * \param[in] exposureModeIndex The index of the current exposure mode
> >> + * \param[in] yHist A Histogram from the ISP statistics to use in constraining
> >> + *	      the calculated gain
> > nit: no indentation
> >
> >> + * \param[in] effectiveExposureValue The EV applied to the frame from which the
> >> + *	      statistics in use derive
> > nit: no indentation
> >
> >> + *
> >> + * Calculate a new exposure value to try to obtain the target. The calculated
> >> + * exposure value is filtered to prevent rapid changes from frame to frame, and
> >> + * divided into shutter time, analogue and digital gain.
> >> + *
> >> + * \return Tuple of shutter time, analogue gain, and digital gain
> >> + */
> >> +std::tuple<utils::Duration, double, double>
> >> +AgcMeanLuminance::calculateNewEv(uint32_t constraintModeIndex,
> >> +				 uint32_t exposureModeIndex,
> >> +				 const Histogram &yHist,
> >> +				 utils::Duration effectiveExposureValue)
> >> +{
> >> +	/*
> >> +	 * The pipeline handler should validate that we have received an allowed
> >> +	 * value for AeExposureMode.
> >> +	 */
> >> +	std::shared_ptr<ExposureModeHelper> exposureModeHelper =
> >> +		exposureModeHelpers_.at(exposureModeIndex);
> >> +
> >> +	double gain = estimateInitialGain();
> >> +	gain = constraintClampGain(constraintModeIndex, yHist, gain);
> >> +
> >> +	/*
> >> +	 * We don't check whether we're already close to the target, because
> >> +	 * even if the effective exposure value is the same as the last frame's
> >> +	 * we could have switched to an exposure mode that would require a new
> >> +	 * pass through the splitExposure() function.
> >> +	 */
> >> +
> >> +	utils::Duration newExposureValue = effectiveExposureValue * gain;
> >> +	utils::Duration maxTotalExposure = exposureModeHelper->maxShutter()
> >> +					   * exposureModeHelper->maxGain();
> >> +	newExposureValue = std::min(newExposureValue, maxTotalExposure);
> >> +
> >> +	/*
> >> +	 * We filter the exposure value to make sure changes are not too jarring
> >> +	 * from frame to frame.
> >> +	 */
> >> +	newExposureValue = filterExposure(newExposureValue);
> >> +
> >> +	frameCount_++;
> >> +	return exposureModeHelper->splitExposure(newExposureValue);
> >> +}
> >> +
> >> +/**
> >> + * \fn AgcMeanLuminance::resetFrameCount()
> >> + * \brief Reset the frame counter
> >> + *
> >> + * This function resets the internal frame counter, which exists to help the
> >> + * algorithm decide whether it should respond instantly or not. The expectation
> >> + * is for derived classes to call this function before each camera start call,
> >> + * either in configure() or queueRequest() if the frame number is zero.
> >> + */
> >> +
> >> +}; /* namespace ipa */
> >> +
> >> +}; /* namespace libcamera */
> >> diff --git a/src/ipa/libipa/agc_mean_luminance.h b/src/ipa/libipa/agc_mean_luminance.h
> >> new file mode 100644
> >> index 00000000..e48dc498
> >> --- /dev/null
> >> +++ b/src/ipa/libipa/agc_mean_luminance.h
> >> @@ -0,0 +1,91 @@
> >> +/* SPDX-License-Identifier: LGPL-2.1-or-later */
> >> +/*
> >> + * Copyright (C) 2024 Ideas on Board Oy
> >> + *
> >> + agc_mean_luminance.h - Base class for mean luminance AGC algorithms
> >> + */
> >> +
> >> +#pragma once
> >> +
> >> +#include <tuple>
> >> +#include <vector>
> >> +
> >> +#include <libcamera/controls.h>
> >> +
> >> +#include "libcamera/internal/yaml_parser.h"
> >> +
> >> +#include "exposure_mode_helper.h"
> >> +#include "histogram.h"
> >> +
> >> +namespace libcamera {
> >> +
> >> +namespace ipa {
> >> +
> >> +class AgcMeanLuminance
> >> +{
> >> +public:
> >> +	AgcMeanLuminance();
> >> +	virtual ~AgcMeanLuminance() = default;
> > There were a few small comments from Laurent that got lost
> >   * destructor in cpp
> 
> 
> The compiler says I can't follow this suggestion; since the estimateLuminance() function is virtual 
> I apparently need a virtual destructor:
> 
> 
> class libcamera::ipa::AgcMeanLuminance’ has virtual functions and accessible non-virtual destructor
> 
> 
> Or am I doing something wrong?

In the .h file,

class AgcMeanLuminance
{
public:
	...
	virtual ~AgcMeanLuminance();
	...
};

In the .cpp file,

AgcMeanLuminance::~AgcMeanLuminance() = default;

> >   * code sytel in enum
> 
> 
> Isn't this addressed by making them lowercase?
> 
> >   * missing line
> >
> > Aside from that, I think we should merge it in.
> >
> > Reviewed-by: Stefan Klug <stefan.klug@ideasonboard.com>
> >
> > Cheers,
> > Stefan
> >
> >> +
> >> +	struct AgcConstraint {
> >> +		enum class Bound {
> >> +			lower = 0,
> >> +			upper = 1
> >> +		};
> >> +		Bound bound;
> >> +		double qLo;
> >> +		double qHi;
> >> +		double yTarget;
> >> +	};
> >> +
> >> +	int parseTuningData(const YamlObject &tuningData);
> >> +
> >> +	void configureExposureModeHelpers(utils::Duration minShutter,
> >> +					  utils::Duration maxShutter,
> >> +					  double minGain,
> >> +					  double maxGain);
> >> +
> >> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes()
> >> +	{
> >> +		return constraintModes_;
> >> +	}
> >> +
> >> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers()
> >> +	{
> >> +		return exposureModeHelpers_;
> >> +	}
> >> +
> >> +	ControlInfoMap::Map controls()
> >> +	{
> >> +		return controls_;
> >> +	}
> >> +
> >> +	double estimateInitialGain();
> >> +	double constraintClampGain(uint32_t constraintModeIndex,
> >> +				   const Histogram &hist,
> >> +				   double gain);
> >> +	utils::Duration filterExposure(utils::Duration exposureValue);
> >> +	std::tuple<utils::Duration, double, double>
> >> +	calculateNewEv(uint32_t constraintModeIndex, uint32_t exposureModeIndex,
> >> +		       const Histogram &yHist, utils::Duration effectiveExposureValue);
> >> +	void resetFrameCount() { frameCount_ = 0; }
> >> +private:
> >> +	virtual double estimateLuminance(const double gain) = 0;
> >> +
> >> +	void parseRelativeLuminanceTarget(const YamlObject &tuningData);
> >> +	void parseConstraint(const YamlObject &modeDict, int32_t id);
> >> +	int parseConstraintModes(const YamlObject &tuningData);
> >> +	int parseExposureModes(const YamlObject &tuningData);
> >> +
> >> +	uint64_t frameCount_;
> >> +	utils::Duration filteredExposure_;
> >> +	double relativeLuminanceTarget_;
> >> +
> >> +	std::map<int32_t, std::vector<AgcConstraint>> constraintModes_;
> >> +	std::map<int32_t, std::shared_ptr<ExposureModeHelper>> exposureModeHelpers_;
> >> +	ControlInfoMap::Map controls_;
> >> +};
> >> +
> >> +}; /* namespace ipa */
> >> +
> >> +}; /* namespace libcamera */
> >> diff --git a/src/ipa/libipa/meson.build b/src/ipa/libipa/meson.build
> >> index 37fbd177..7ce885da 100644
> >> --- a/src/ipa/libipa/meson.build
> >> +++ b/src/ipa/libipa/meson.build
> >> @@ -1,6 +1,7 @@
> >>   # SPDX-License-Identifier: CC0-1.0
> >>   
> >>   libipa_headers = files([
> >> +    'agc_mean_luminance.h',
> >>       'algorithm.h',
> >>       'camera_sensor_helper.h',
> >>       'exposure_mode_helper.h',
> >> @@ -10,6 +11,7 @@ libipa_headers = files([
> >>   ])
> >>   
> >>   libipa_sources = files([
> >> +    'agc_mean_luminance.cpp',
> >>       'algorithm.cpp',
> >>       'camera_sensor_helper.cpp',
> >>       'exposure_mode_helper.cpp',