[libcamera-devel,7/8] libcamera: v4l2_videodevice: Map V4L2_PIX_FMT_GREY to DRM FourCC
diff mbox series

Message ID 20200313233856.25202-8-laurent.pinchart@ideasonboard.com
State Superseded
Headers show
Series
  • Simple pipeline handler
Related show

Commit Message

Laurent Pinchart March 13, 2020, 11:38 p.m. UTC 
DRM has a format for 8-bit greyscale data, DRM_FORMAT_R8. Despite the
'R' name, which comes from GL/Vulkan to mean single-channel data, the
format maps to greyscale for display. We can thus map it to
V4L2_PIX_FMT_GREY.

Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
---
 src/libcamera/v4l2_videodevice.cpp | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

Comments

Nicolas Dufresne March 14, 2020, 11:14 a.m. UTC | #1 
Le samedi 14 mars 2020 à 01:38 +0200, Laurent Pinchart a écrit :
> DRM has a format for 8-bit greyscale data, DRM_FORMAT_R8. Despite the
> 'R' name, which comes from GL/Vulkan to mean single-channel data, the
> format maps to greyscale for display. We can thus map it to
> V4L2_PIX_FMT_GREY.

The mapping looks fine to me, but may lead to surprising result in
absence of colorimetry support. Most YCbCr 4:0:0 will be limited range,
while on the GFX side, most of R8 usage will be full range. Just be
aware.

Nicolas

> 
> Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> ---
>  src/libcamera/v4l2_videodevice.cpp | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)
> 
> diff --git a/src/libcamera/v4l2_videodevice.cpp b/src/libcamera/v4l2_videodevice.cpp
> index f68b67b33293..f89bf2ff781e 100644
> --- a/src/libcamera/v4l2_videodevice.cpp
> +++ b/src/libcamera/v4l2_videodevice.cpp
> @@ -1473,12 +1473,15 @@ PixelFormat V4L2VideoDevice::toPixelFormat(uint32_t v4l2Fourcc)
>  	case V4L2_PIX_FMT_NV21M:
>  		return DRM_FORMAT_NV21;
>  
> +	/* Greyscale formats. */
> +	case V4L2_PIX_FMT_GREY:
> +		return DRM_FORMAT_R8;
> +
>  	/* Compressed formats. */
>  	case V4L2_PIX_FMT_MJPEG:
>  		return DRM_FORMAT_MJPEG;
>  
>  	/* V4L2 formats not yet supported by DRM. */
> -	case V4L2_PIX_FMT_GREY:
>  	default:
>  		/*
>  		 * \todo We can't use LOG() in a static method of a Loggable
> @@ -1557,6 +1560,10 @@ uint32_t V4L2VideoDevice::toV4L2Fourcc(PixelFormat pixelFormat, bool multiplanar
>  	case DRM_FORMAT_NV21:
>  		return V4L2_PIX_FMT_NV21;
>  
> +	/* Greyscale formats. */
> +	case DRM_FORMAT_R8:
> +		return V4L2_PIX_FMT_GREY;
> +
>  	/* Compressed formats. */
>  	case DRM_FORMAT_MJPEG:
>  		return V4L2_PIX_FMT_MJPEG;
Laurent Pinchart March 14, 2020, 11:47 a.m. UTC | #2 
Hi Nicolas,

On Sat, Mar 14, 2020 at 07:14:12AM -0400, Nicolas Dufresne wrote:
> Le samedi 14 mars 2020 à 01:38 +0200, Laurent Pinchart a écrit :
> > DRM has a format for 8-bit greyscale data, DRM_FORMAT_R8. Despite the
> > 'R' name, which comes from GL/Vulkan to mean single-channel data, the
> > format maps to greyscale for display. We can thus map it to
> > V4L2_PIX_FMT_GREY.
> 
> The mapping looks fine to me, but may lead to surprising result in
> absence of colorimetry support. Most YCbCr 4:0:0 will be limited range,
> while on the GFX side, most of R8 usage will be full range. Just be
> aware.

Thanks for the comment. In the specific case I'm trying to support, data
comes from a sensor that doesn't have a Bayer filter, so it's really
full range greyscale (it's actually 10 bits scaled down to 8 bits). In
the general case, we need to convey colorimetry information, I agree
with you, and that's something that can't be conveyed just through a
4CC, we will need a dedicated API.

> > Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
> > ---
> >  src/libcamera/v4l2_videodevice.cpp | 9 ++++++++-
> >  1 file changed, 8 insertions(+), 1 deletion(-)
> > 
> > diff --git a/src/libcamera/v4l2_videodevice.cpp b/src/libcamera/v4l2_videodevice.cpp
> > index f68b67b33293..f89bf2ff781e 100644
> > --- a/src/libcamera/v4l2_videodevice.cpp
> > +++ b/src/libcamera/v4l2_videodevice.cpp
> > @@ -1473,12 +1473,15 @@ PixelFormat V4L2VideoDevice::toPixelFormat(uint32_t v4l2Fourcc)
> >  	case V4L2_PIX_FMT_NV21M:
> >  		return DRM_FORMAT_NV21;
> >  
> > +	/* Greyscale formats. */
> > +	case V4L2_PIX_FMT_GREY:
> > +		return DRM_FORMAT_R8;
> > +
> >  	/* Compressed formats. */
> >  	case V4L2_PIX_FMT_MJPEG:
> >  		return DRM_FORMAT_MJPEG;
> >  
> >  	/* V4L2 formats not yet supported by DRM. */
> > -	case V4L2_PIX_FMT_GREY:
> >  	default:
> >  		/*
> >  		 * \todo We can't use LOG() in a static method of a Loggable
> > @@ -1557,6 +1560,10 @@ uint32_t V4L2VideoDevice::toV4L2Fourcc(PixelFormat pixelFormat, bool multiplanar
> >  	case DRM_FORMAT_NV21:
> >  		return V4L2_PIX_FMT_NV21;
> >  
> > +	/* Greyscale formats. */
> > +	case DRM_FORMAT_R8:
> > +		return V4L2_PIX_FMT_GREY;
> > +
> >  	/* Compressed formats. */
> >  	case DRM_FORMAT_MJPEG:
> >  		return V4L2_PIX_FMT_MJPEG;

Patch
diff mbox series 

diff --git a/src/libcamera/v4l2_videodevice.cpp b/src/libcamera/v4l2_videodevice.cpp
index f68b67b33293..f89bf2ff781e 100644
--- a/src/libcamera/v4l2_videodevice.cpp
+++ b/src/libcamera/v4l2_videodevice.cpp
@@ -1473,12 +1473,15 @@  PixelFormat V4L2VideoDevice::toPixelFormat(uint32_t v4l2Fourcc)
 	case V4L2_PIX_FMT_NV21M:
 		return DRM_FORMAT_NV21;
 
+	/* Greyscale formats. */
+	case V4L2_PIX_FMT_GREY:
+		return DRM_FORMAT_R8;
+
 	/* Compressed formats. */
 	case V4L2_PIX_FMT_MJPEG:
 		return DRM_FORMAT_MJPEG;
 
 	/* V4L2 formats not yet supported by DRM. */
-	case V4L2_PIX_FMT_GREY:
 	default:
 		/*
 		 * \todo We can't use LOG() in a static method of a Loggable
@@ -1557,6 +1560,10 @@  uint32_t V4L2VideoDevice::toV4L2Fourcc(PixelFormat pixelFormat, bool multiplanar
 	case DRM_FORMAT_NV21:
 		return V4L2_PIX_FMT_NV21;
 
+	/* Greyscale formats. */
+	case DRM_FORMAT_R8:
+		return V4L2_PIX_FMT_GREY;
+
 	/* Compressed formats. */
 	case DRM_FORMAT_MJPEG:
 		return V4L2_PIX_FMT_MJPEG;