[x264-devel] Custom slice support

[List, Peter]

Dear Mojtaba,

> On Tue, 2007-11-13 at 17:49 +0100, List, Peter wrote:
> 
> > 1. there is no prediction over slice boundaries
> 
> Yes. I believe x264 already takes care of not predicting from
> mbs in another slice.

Yes, it does, but only if you HAVE ACTIVATED SLICES, which is not
supported anymore.


> 
> > 2. EACH of the slice-headers contain ALL the information that was
> >    formally called "Frame-header"
> >
> Each slice begins with the slice header and NAL start code, so
> the decoder can resync to the NAL start code if the previous slice
> is lost or damaged. I believe it is not necessary to have frame header
>  for each slice. Slice header should suffice.

In H.264 "frame headers" DO NOT exist.
Also: if the first slice is lost, how could the decoder know which
parameters this frame uses? It can only assume, that the same parameters
are used as in the last frame (which might be wrong though).

> > 3. It also contains the address where it starts, so it can always be
> >    placed at the right spot in the frame (even if slices at the top
> >    of that frame are missing.
> >
> This is part of slice header (start mb of each slice is in its
header).

Yes, it does, but only if you HAVE ACTIVATED SLICES, which is not
supported anymore.




>>> Hello Peter,
>>>   Have you tried the patch I sent to this mailing list that adds 
>>> splitting each frame into NALs when its size is greater than a given

>>> programmable threshold?
>>>
>>>
http://mailman.videolan.org/pipermail/x264-devel/2007-October/003694.htm
l
>>>
>>> I have been using that patch for our h264 transmission over lossy 
>>> networks. It would be good to know if others use it and report any 
>>> problems so I can fix it. Others using it would also give x264 
>>> contributors more confidence before merging the patch.
>>>
>>> Mojtaba Hosseini