[XviD-devel] I have trouble with the B-frame's standard

[Christoph Lampert]

On Wed, 24 Jul 2002, Michael Niedermayer wrote:

> Hi
> 
> On Wednesday 24 July 2002 20:16, Christoph Lampert wrote:
> > Hi,
> >
> > either I am too stupid or I simply can't read. Maybe one of you
> > can help me, if he has a copy of the standard of a draft.
> >
> > If I encode a B-frame's macroblock and the following P-frame's macroblock
> > at the same position ("co-located block") is SKIPed, the B-frame MB is
> > automatically encoded as FORWARD with (0,0) motion vector and cbp==0,
> > so it's also just a copy of the very first MB. In a sequence PBBP all
> > macroblocks at this position would be identical. Am I right?
> yes
> 
> >
> > I don't want this! 
> me neither ;) and i doubt anyone except iso-mpeg does like it
> 
> > I want to code the B-frames MBs individually, in
> > particular they are NOT supposed to be a copy, there is an object at this
> > position (only in the B, not in the surroundig P-frames).
> > I want to code a motion vector, or at least have cbp!=0
> >
> > What should I do?
> well, there isnt much choice, u have to check if the MBs in the b frames 
> between p frames have any significantly different MBs and than force the 
> colocated MBs of these in the next P frame to be NOT skiped, see ffmpeg CVS 
> h263.c line 398 
> 
> [...]


THIS SUCKS!!!!

And I say it again...  IT SUCKS!!!!!!!!!!!!!!!!!!!!

Boy, didn't these people _test_ the standard before making it one?
I have spend several hours finding where I was wrong, because nobody could
be so stupid and now it come out it's simply a "bug" in the standard? 

I don't even _have access_ to the intermediate b-frames when coding the
following P-frame. :-( 

gruel

-- 
Christoph H. Lampert chl@math.uni-bonn.de | Diese Signature wurde maschi-     
Beringstr. 6, Raum 14 Tel. (0228) 73-2948 | nell erstellt und bedarf
Sprechstunden: keine, aber meistens da    | keiner Unterschrift. AZ 27B-6