[FFmpeg-trac] #7686(avfilter:new): Wrong decision of pixel format in filtergraph

[FFmpeg]

#7686: Wrong decision of pixel format in filtergraph
------------------------------------+------------------------------------
             Reporter:  mfcc64      |                    Owner:
                 Type:  defect      |                   Status:  new
             Priority:  normal      |                Component:  avfilter
              Version:  git-master  |               Resolution:
             Keywords:              |               Blocked By:
             Blocking:              |  Reproduced by developer:  0
Analyzed by developer:  0           |
------------------------------------+------------------------------------

Comment (by mfcc64):

 OK,
 for lutyuv:
 actual: rgb24 -> yuv444p -> yuv444p12be
 ideal: rgb24 -> yuv444p12le -> yuv444p12be
 The ideal pixfmt should be yuv444p12le because it is exactly yuv444p12be
 with different endianness. (OK, one can argue that yuv444p is closer to
 rgb24).

 for showcqt:
 actual: none -> yuv420p -> bgr24
 ideal: none -> rgb24 -> bgr24
 The ideal pixfmt should be rgb24 because it is exactly bgr24 with
 different red-green-blue ordering. And no one can argue that yuv420p is
 closer to 'none'.

 The root cause as you stated is similar. The pixfmt decision on filter is
 solely based on input filters when there are no common pixfmts.

 I posted lutyuv example because I was retrying to fix
 https://ffmpeg.org/pipermail/ffmpeg-devel/2016-June/195941.html

--
Ticket URL: <https://trac.ffmpeg.org/ticket/7686#comment:4>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker