[FFmpeg-devel] [PATCH] lavf/utils: Fix DTS for short H264 streams.

Sasi Inguva isasi at google.com
Fri Mar 11 22:54:07 CET 2016


For testing , a link to one such failing video
https://drive.google.com/file/d/0Bz6XfEJZ-9N3X0FSMWR6T3B3U28/view?usp=sharing
 . Without this patch,

ffprobe -show_frames -print_format compact  | grep media_type=video  -
shows dts=N/A

On Fri, Mar 11, 2016 at 12:39 PM, Sasi Inguva <isasi at google.com> wrote:

> With this patch it doesn't produce invalid DTS for the file you mentioned
> tickets/1242/sample.mkv
> ./ffmpeg -i sample.mkv -c:v copy -an -t 1 -f framecrc -
>
> On Fri, Mar 11, 2016 at 12:36 PM, Sasi Inguva <isasi at google.com> wrote:
>
>> Having another go at this. My goal is to fix DTS of H264 videos which are
>> shorter than 7 frames and don't have B-frames. Currently ffmpeg will output
>> DTS=N/A for such frames.
>>
>> On Fri, Mar 11, 2016 at 12:34 PM, Sasi Inguva <isasi at google.com> wrote:
>>
>>> Fill DTS if all packets have been read in avformat_find_stream_info, and
>>> still
>>> has_decode_delay_been_guessed returns false.
>>>
>>> Signed-off-by: Sasi Inguva <isasi at google.com>
>>> ---
>>>  libavformat/utils.c | 37 +++++++++++++++++++++++++++++++++++++
>>>  1 file changed, 37 insertions(+)
>>>
>>> diff --git a/libavformat/utils.c b/libavformat/utils.c
>>> index 5f48de1..f22309d 100644
>>> --- a/libavformat/utils.c
>>> +++ b/libavformat/utils.c
>>> @@ -877,6 +877,7 @@ static int has_decode_delay_been_guessed(AVStream
>>> *st)
>>>         avpriv_h264_has_num_reorder_frames(st->codec) ==
>>> st->codec->has_b_frames)
>>>          return 1;
>>>  #endif
>>> +
>>>      if (st->codec->has_b_frames<3)
>>>          return st->nb_decoded_frames >= 7;
>>>      else if (st->codec->has_b_frames<4)
>>> @@ -3165,6 +3166,7 @@ int avformat_find_stream_info(AVFormatContext *ic,
>>> AVDictionary **options)
>>>      int64_t max_stream_analyze_duration;
>>>      int64_t max_subtitle_analyze_duration;
>>>      int64_t probesize = ic->probesize;
>>> +    int eof_reached = 0;
>>>
>>>      flush_codecs = probesize > 0;
>>>
>>> @@ -3331,6 +3333,7 @@ int avformat_find_stream_info(AVFormatContext *ic,
>>> AVDictionary **options)
>>>
>>>          if (ret < 0) {
>>>              /* EOF or error*/
>>> +            eof_reached = 1;
>>>              break;
>>>          }
>>>
>>> @@ -3454,6 +3457,40 @@ int avformat_find_stream_info(AVFormatContext
>>> *ic, AVDictionary **options)
>>>          count++;
>>>      }
>>>
>>> +    if (eof_reached && ic->internal->packet_buffer) {
>>> +        int stream_index;
>>> +        for (stream_index = 0; stream_index < ic->nb_streams;
>>> stream_index++) {
>>> +            AVPacketList *pktl = ic->internal->packet_buffer;
>>> +            int64_t pts_buffer[MAX_REORDER_DELAY+1];
>>> +
>>> +            // EOF already reached while reading the stream above.
>>> +            // So continue with reoordering DTS with whatever delay we
>>> have.
>>> +            int codec_delay = st->codec->has_b_frames;
>>> +
>>> +            st = ic->streams[stream_index];
>>> +
>>> +            if (st->codec->codec_id != AV_CODEC_ID_H264 ||
>>> +                has_decode_delay_been_guessed(st)) {
>>> +                continue;
>>> +            }
>>> +
>>> +
>>> +            for (; pktl; pktl = get_next_pkt(ic, st, pktl)) {
>>> +                if (pktl->pkt.stream_index != stream_index)
>>> +                    continue;
>>> +
>>> +                if (pktl->pkt.pts != AV_NOPTS_VALUE && codec_delay <=
>>> MAX_REORDER_DELAY) {
>>> +                    int j;
>>> +                    pts_buffer[0] = pktl->pkt.pts;
>>> +                    for (j = 0; j < codec_delay && pts_buffer[j] >
>>> pts_buffer[j + 1]; j++)
>>> +                      FFSWAP(int64_t, pts_buffer[j], pts_buffer[j + 1]);
>>> +
>>> +                    pktl->pkt.dts = select_from_pts_buffer(st,
>>> pts_buffer, pktl->pkt.dts);
>>> +                }
>>> +            }
>>> +        }
>>> +    }
>>> +
>>>      if (flush_codecs) {
>>>          AVPacket empty_pkt = { 0 };
>>>          int err = 0;
>>> --
>>> 2.7.0.rc3.207.g0ac5344
>>>
>>>
>>
>


More information about the ffmpeg-devel mailing list