[MPlayer-dev-eng] XviD Vbr ready for use?

Christoph Lampert chl at math.uni-bonn.de
Wed Sep 4 08:58:04 CEST 2002


On Tue, 3 Sep 2002 mplayer-dev-eng-request at mplayerhq.hu wrote:
> Now that is interesting. Diego asked for volunteers some days ago to
> reimplement it to solve these licensing issues. I decided to have
> a deep look into it and try to do the reimplementation. 
>
> Do you think this pays off, or will the library be ready for use 
> soon instead?

I can't really tell. As you might have heard, XVID has serious problems
with Sigma Designs at the moment and development has been stopped for
"political" reasons. 
Current version _is_ already better than OpenDivX 2-pass and much more
flexible, but because of bad parameters the keyframes tend to get too bad
quality. The same code is used in XVID VfW API successfully for several
month. The Sigma stuff came at a bad time, otherwise there surely would
be a "good" version already. 

About parameter optimization: Of course the author will look for good
parameters for XviD, but I doubt that the difference in two-pass
behaviour will be that big between XviD and DivX4/5 or even ffmpeg. 
It's rather a question of how to detect difficult frames. 

> Can you tell me which parts in particular are subject to the 
> OpenDivX license?

I'm not sure. The original was 2-pass implementation by Sparky (if I
remember correctly). I modified it, but I don't know if the modifications
are large enough to consider it "new work". Maybe almost the whole file is
still OpenDivX. I rather adapted the function calls etc. not reimplemented
anything. 

It would be safest to study the way it is done and then write the whole
thing from scratch, but maybe it's not worth to create two implementations
of two-pass, when there is one nearly finished library where only
parameters have to be adapted to different codecs.

Christoph 




More information about the MPlayer-dev-eng mailing list