Re: [XviD-devel] Adaptive quantisation

[Dirk Knop <dknop@xxxxxxx>]

Hi,

Michael Militzer wrote:

> well, I don't know. Pretty much thresholds are needed. But I think I'll add
> some additional code (activity masking), after I looked at the adapt quant
> code again anyway, this needs no API change (flags). So you'll all get
> something new to play with ;-)
Well, for now I'd be satisfied if I got the old luma masking code back 
in CVS since this really seems to be important to the users.

> btw: there's a new codec comparison? I didn't knew this too. From what you
> said, we're not very good this time?
>  
XviD got even disqualified for SPR as the scenes tested were a huge 
"block-fest":-(
Well, I spent several hours explaining Doom9 how to use XviD, he did 
several new encodes (he listened to another person as well [namely 
MaTTer] and mixed what he thought was best *grrr*). Maybe this is the 
effect of the luma masking code as it is now like you explained in the 
first reply. I think the old luma masking code would help those scenes.
On the other hand, Doom9 has no interest in those codecs anymore, but in 
DVD authoring, so it's clear that he doesn't like to fiddle around with 
a codec but wants some results with some easy settings (well, back in 
the nandub times this was way different...).

Sorry that the link I came up with doesn't contain the real useful 
information, but it all started there at least.

Regards,
Dirk