[XviD-devel] Comparison "old" vs "new" adaptive quant code

[Christoph Lampert]

> >Sounds like an error, indeed, maybe connected to adaptive quantization? 
> >EPZS is (if at all) not tested with combinations like adaptive quant,
> >modulated quant etc.
> >
> Still, I'm looking forward when that is finished, as it (see above) does 
> some magic on details. But so long I'll stick with PMVfast, of course. 
> Btw., is it safe to use PMV_USESQUARES16 with that as default for 
> "search precision 6" as well? Is the gain which EPZS shows deriving from 
> there? Sorry, this is a dumb question, but I'm just a little too weak to 
> understand the implementation right, it gives me a headache every time I 
> try to read that :)

USE_SQUARES simply searched more points in every step (a 3x3 square of 8
neighbours instead of a diamond with 4 neigbours). So it's slower, but
also less likely to get trapped in a local minimum. It's save to use 
USE_SQUARES with PMVfast. 

Btw. I don't know if it's default, but you can also use ADVANCED_DIAMOND
instead. That's somehow a more clever way of using diamonds. It could
be even faster that normal diamond and still give quality similar to
SQUARES. 

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