[FFmpeg-devel] [PATCH 13/18] h264_sei: parse the picture timing SEIs correctly

[Anton Khirnov]

Quoting Kieran Kunhya (2020-03-23 22:34:59)
> >
> > I don't think you read that section carefully enough. What it says (in
> > my interpretation) is that an SPS can be activated only by:
> > - buffering period SEI
> > - a PPS (which is itself activated by picture data)
> > Picture timing SEI is neither of those.
> >
> 
> The contents of pic-struct do not affect activation but have dependent
> syntax elements.
> So for example a Buffering Period SEI could activate an SPS and then
> followed by pic-struct and pic struct would then have a dependency on the
> activated SPS.
> This SPS could be different to the one the VCL data refers to. I think
> there are other corner cases as well.

I don't think it could be different. The active SPS must not change in a
sequence and the SEI messages are all a part of the same sequence as the
coded pictures the belong to. So even if there was an buffering period
SEI that activates some SPS, it has to be the same SPS that will be used
for decoding the slices.

-- 
Anton Khirnov