[vlc-devel] [vlc-commits] commit: Support H.264 SEI recovery points. (Steinar H. Gunderson )

[Laurent Aimar]

Hi,

On Sat, Oct 02, 2010 at 12:10:56AM +0200, Steinar H. Gunderson wrote:
> On Sat, Oct 02, 2010 at 12:03:28AM +0200, Laurent Aimar wrote:
> > Now unless I am wrong, the SEI recovery point specify recovery_frame_cnt in frames (and
> > not in pictures) and in the presentation orders (not the decoding order).
> 
> Care to explain? What's the difference between frames and pictures?
A picture represents a field or a frame. A frame is composed of two fields.
And yes, some of the fields decoder_sys_t::*frame* are misnamed in that
regards :(

> I cannot comment on presentation vs. decoder order, since I don't have a copy
> of the standard in front of me. B-frames + PIR is a bit of a weird
> combination, but sure, I guess it should be supported.

"recovery_frame_cnt specifies the recovery point of output pictures in output order. All decoded pictures in output order
are indicated to be correct or approximately correct in content starting at the output order position of the reference picture
having the frame_num equal to the frame_num of the VCL NAL units for the current access unit incremented by
recovery_frame_cnt in modulo MaxFrameNum arithmetic. recovery_frame_cnt shall be in the range of 0 to
MaxFrameNum - 1, inclusive."

 So it seems it is even a bit more complex than just counting the number of
frames.

 The thing is that the SEI recovery point can be and is used without PIR, and
so I am not sure of the consequences to incorrectly handle it for those
cases. Besides, IMHO in the long term, it's better to not have a feature than
having a partial 'kludge'.

Regards,

-- 
fenrir