[x264-devel] about single layer video coding and scalable video coding

[Jason Garrett-Glaser]

On Fri, Jun 19, 2009 at 12:24 AM, Loic Le Loarer<lll+vlc at m4x.org> wrote:
> Hi,
>
> On Thursday June 18 2009 at 12:30:26 PM +0200, 武南龚 wrote:
>> thank you so much. I'm a little puzzled by the sentence below, can you give
>> me some explanation.
>>
>> Yes, temporal scalability is just a single layer with non-reference
>> frames at regular intervals.
>> _______________________________________
>>
>> besides, x264 doesn't support SVC, can I treat the single layer coded video
>> as a temporal scalable video coded one?
>
> Yes, you can do that if the coded stream has the correct properties. For
> example, do be a correct temporal scalable video sequence with 2 layers,
> the base layer at 15fps and the second at 30fps, it is enough to have
> the following sequence of pictures in display order (using old MPEG2
> style for references):
>
> I B P B P B P...
>
> The base layer contains only the I and P pictures, the second layer adds
> the B pictures.
>
> Using x264, it is possible to generate such a sequence using the
> following options (or similar):
>
>  --bframes 1 --b-adapt 0 --ref 2
>
> Using the --b-pyramid option, it is also probably possible to generate
> a 3 layers stream, the first layer containning I/P frames, the second
> reference B frames and the third non-reference B frames.

This is not entirely true.  x264 adds the pyramidal (referenced)
B-frame to the reference list rather than disposing of it with MMCO,
which means that b-pyramid does *not* add an extra layer of temporal
scalability.

Patches to add MMCO support and resolve this are welcome.

Dark Shikari