[Ffmpeg-cvslog] CVS: ffmpeg/libavcodec snow.c, 1.68, 1.69 avcodec.h, 1.436, 1.437

[Loren Merritt]

On Wed, 28 Dec 2005, Jeff Clagg wrote:

> On Wed, Dec 28, 2005 at 03:32:11PM -0800, Corey Hickey wrote:
>
>> Besides that, what is *cmp=12? It doesn't seem to be documented.
>
> Quoth mplayer's DOCS/tech/snow.txt:
>
> * cmp, subcmp, mbcmp
>   Set the comparison function, default: 0 (SAD).
>   useful values = 0 (SAD), 1 (SSD), 11 (5/3 wavelet), 12 (9/7 wavelet).
>   Experience shows that SSD is the best most of the time, while SAD is
>   slightly better the remainder of the time.
>
> [snip]
>
> Then, at the end of that document:
>
>   v4mv and the wavelet cmp functions are theoretically good, but in
>   practice won't work well until we get OBMC-aware motion estimation.
>
>   In short:
>   The best options in almost all cases are
>   vcodec=snow:vstrict=-2:vqscale=$N:pred=0:cmp=1:subcmp=1:mbcmp=1:qpel

In particular, I hypothesized that v4mv would require obmc-aware block 
size decision, and that wavelet cmp funcs would require comparing the 
whole area of support at once. Neither of those has been implemented yet.

--Loren Merritt