[XviD-devel] Early SKIP and early DC-only

[Christoph Lampert]

On Mon, 21 Apr 2003, Marco Al wrote:

> From: "Christoph Lampert" <chl at math.uni-bonn.de>
> 
> > > hard to comprehend how MPEG4 ended up with its braindead scheme.
> >
> > Hm, I didn't find it that bad. What do you dislike about it?
> 
> In p-frames the skip decision in MPEG4 is tied to the coding of backgrounds
> in static camera and animation shots ... and can sometimes be abused for
> coding different things

Background, yes. But with GMC it's not static camera anymore. I find this
rather logical: If there is no global motion (at least none is specified)
SKIP means (,0,0), otherwise, it means the global MV. 
 
> The skip decision should be tied to the predictibility of a MB, not tied to
> specific forms of video ... so that means not taking 0,0 as the MV of a
> skipped MB, but the predictor (like H.264, which AFAIK defaults to using the
> predictor as the MV for skipped blocks ... only using 0,0 if any of the MVs
> used for prediction is 0,0).

On the other hand, you have 2 choices of encoding with few bits then:
a) SKIP, automatic MV (0,0)  1 bit
b) cbp==0, pmv=0    which is what you would like as SKIP  (6 bits?)

5 bits per MB in the worst case, that's not too much of a difference....

gruel