[FFmpeg-devel] H.264 MVC

[Stefan Gehrer]

On 06/25/2010 07:18 PM, Peter Wimmer wrote:
> Hi,
>
>> IMHO it would be easiest to decode into side-by-side format,
>> the output can then just be handled like other 3d content.
>
> Over/under would be smarter, because each view would be located in two
> contiguous memory blocks, while data would have to be interleaved for
> side-by-side.

In ffmpeg as well as any sane software a rectangular image
(or plane thereof) is represented by pointer, width, height
and a stride (address distance from one line to the next).
In this representation, side-by-side can be separated into
the two views just as easily as top-bottom without any
copying/deinterleaving.
Same goes for the decoders: the existing H.264 decoder could
decode a frame 1920 pixels wide into a buffer 3840 pixels wide,
no code change needed.
So i think side-by-side is just as good in the sense of
memory organisation as top-bottom and seems to be used more
widely.

Stefan