| Message ID | 20261029132631.370507-1-halfline@gmail.com |
|---|---|
| Headers | show |
| Series |
|
| Related | show |
Hi Ray Quoting Ray Strode (2026-10-29 13:26:22) > Hi, > > Here's an updated version of the software ISP crop series, rebased on > current master (6e19ca0f). > I don't know if this is specific to my client or circumstances, but when you post new versions as a reply to the existing version - the thread is already marked as superceeded, so it means I don't process new versions. -- Kieran > It makes GPU statistics follow the sensor region visible in the output, > keeping exposure and white balance metering aligned with the image. The > simple pipeline also reports the configured CPU or GPU crop through > ScalerCropMaximum and ScalerCrop request metadata. > > The series depends on Martin's GPU crop-centering patch [1] and the > accompanying getSelection() API patch. > > I've implemented the optional-result approach from the earlier > getSelection() discussion as a separate patch. Success with an empty > optional means no rectangle was reported for the current configuration. > > Following Stefan's feedback, the crop coordinates are relative to the > physical pixel array. The first patch corrects the ScalerCrop and > ScalerCropMaximum documentation to match. > > [1] https://patchwork.libcamera.org/patch/28178/ > > --Ray > > Ray Strode (9): > libcamera: Document ScalerCrop in physical pixel-array coordinates > software-isp: Align statistics with the visible sensor crop > tests: Validate the visible sensor crop > software-isp: Expose the configured image crop > tests: Check the configured CPU image crop > software-isp: Map input crops to native sensor coordinates > tests: Validate native sensor crop coordinates > pipeline: simple: Report the software ISP's maximum crop > pipeline: simple: Include the software ISP crop in request metadata > > .../internal/software_isp/debayer_geometry.h | 48 ++++++++++ > .../libcamera/internal/software_isp/meson.build | 1 + > .../libcamera/internal/software_isp/software_isp.h | 1 + > src/libcamera/control_ids_core.yaml | 7 +- > src/libcamera/pipeline/simple/simple.cpp | 95 +++++++++++++++++- > src/libcamera/property_ids_core.yaml | 4 +- > src/libcamera/software_isp/debayer.cpp | 11 +++ > src/libcamera/software_isp/debayer.h | 1 + > src/libcamera/software_isp/debayer_cpu.h | 4 + > src/libcamera/software_isp/debayer_egl.cpp | 15 +-- > src/libcamera/software_isp/debayer_egl.h | 3 +- > src/libcamera/software_isp/software_isp.cpp | 14 +++ > test/debayer-geometry.cpp | 106 +++++++++++++++++++++ > test/meson.build | 2 + > test/software_isp/debayer-crop.cpp | 102 ++++++++++++++++++++ > test/software_isp/meson.build | 20 ++++ > 16 files changed, 414 insertions(+), 20 deletions(-)
Hi, On Thu, Oct 29, 2026 at 9:33 AM Kieran Bingham <kieran.bingham@ideasonboard.com> wrote: > I don't know if this is specific to my client or circumstances, but when > you post new versions as a reply to the existing version - the thread is > already marked as superceeded, so it means I don't process new versions. Okay, i'll post a v5 as a new thread, that will give me a chance to do another pass over the commit messages anyway. --Ray