[FFmpeg-devel] [PATCH] avcodec/h264dec: add option to ignore in band parameter set

["zhilizhao(赵志立)"]

> On Feb 6, 2022, at 11:31 PM, Hendrik Leppkes <h.leppkes at gmail.com> wrote:
> 
> On Sun, Feb 6, 2022 at 3:58 PM Zhao Zhili <quinkblack at foxmail.com> wrote:
>> 
>> It works in three mode:
>> 0: don't ignore in-band ps
>> 1: ignore in-band ps
>> -1: if corrupted data is detected, then ignore in-band ps afterwards
>> 
>> h264dec working hard to do error resilience, it doesn't drop a
>> whole packet when error is detected in a nalu. Then there is a
>> higher chance for fake sps/pps be found and used. This happened
>> in a mp4 file. h264dec failed to recovery after broken data due
>> to the fake pps, while other H.264 decoders have no such problem.
> 
> This seems extremely hacky. I think you need to elaborate on the
> problem you are trying to fix, so that any approach to a solution can
> be judged properly.

There are two problems here:

Firstly, after split a packet into multiple NALUs, if part of these NALUs
have broken data, drop the whole packet or not. For H.264 with avcc
bitstream, the split itself is questionable if part of these NALUs are
invalid. Keep going after broken H264_NAL_SLICE has no serious effect,
but updating SPS/PPS in this case can be dangerous.

Secondly, for some containers like mp4, updating SPS/PPS are supported by
spec but don’t get widely support among implementations. FFmpeg can
generate mp4 with SPS/PPS updating in-band, but demux and decoding only
works without seek. On the whole, it make sense to have an option do
ignore in-band parameter set for those container.

> 
> - Hendrik
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> 
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".