Closed frankplow closed 2 months ago
the file is a subpic file. I checked spec, subpic may be a little hard for gsoc, I will try to pick up your https://github.com/ffvvc/FFmpeg/pull/191 and work on it.
@nuomi2021 Yes I worked this is due to subpics also, sorry should have posted here. I think #191 is all the derivations etc. for pps_single_slice_per_subpic_flag
/pps_rect_slice_flag
. I think the errors in #191 as it stands are the same errors we see in LMCS_B_Dolby and the other subpic bitstreams and are because we don't handle sps_independent_subpics_flag
. Unfortunately it doesn't seem there are any test bitstreams with pps_single_slice_per_subpic_flag=1
and sps_independent_subpics_flag=0
, so it is difficult to test seperately.
fixed by https://github.com/FFmpeg/FFmpeg/commit/238bb653e7ea466a024d027072b57d0a9a1278d7, thank you @frankplow
The conformance bitstream LMCS_B_Dolby is currently failing. At the moment, this is due to an AVERROR_INVALIDDATA. This can be resolved with this patch: https://patchwork.ffmpeg.org/project/ffmpeg/patch/20240126162043.59687-1-post@frankplowman.com/
Even with that patch applied however, the test still fails. YUView shows the first frames are identical, but all subsequent frames differ. Together with the visual nature of the distortions, inter prediction looks suspect.