View Full Version : Custom quantizer matrix
little_buddha
27th April 2004, 10:53
Hi
I've found some posts talking about the quantizer matrix and the different coefficients in it. I was searching some examples of custom matrix... I would like to keep more details in dark foggy scenes for example (i know the file size will be bigger).
Does someone know where i can find informations ?
tchuss
mikeX
27th April 2004, 14:40
Yes, on that little button that says 'search' : P
Firts you might wanna take a look at the Unofficial XviD FAQ (www.vslcatena.nl/~ronald/docs/xvidfaq.html)
Then you can start trying out some custom matrices for yourself, the XviD binaries provided by Koepi come with quite a collection of them.
You can also look at some of the custom QM specific threads here, there are quite a few of them I believe...
kilg0r3
28th April 2004, 07:55
This specific problem cannot be reolved through a custom matrix. It is a problem inherent to the current implementation of xvid. IIRC syskin said that there will be a fix in the 1.1 version.
little_buddha
28th April 2004, 09:31
ok thanks for answering.
i knew there are some posts dealing with quantizer matrices in doom9 and the unofficial xvid faq, but i was searching for examples because i don't know the "reasonable" values for the coefficients (perhaps i've not searched enough :rolleyes: ).
This specific problem cannot be reolved through a custom matrix. It is a problem inherent to the current implementation of xvid. IIRC syskin said that there will be a fix in the 1.1 version.
I think this problem is partially related to b-frames. Disabling b-frames seems to reduce those "crappy" DCT blocks in dark scrolling scenes. Some attempts with custom matrices (and b-frames enabled) let me think that it can influence. The solution is probably to assign a higher bitrate on these scenes, either in xvid options (quantizer matrix ???) or in avisynth script by adding noise.
kilg0r3
28th April 2004, 11:29
Ha! Found the thread (http://forum.doom9.org/showthread.php?s=&threadid=70360&highlight=dark%2A+edge%2A)
little_buddha
28th April 2004, 11:45
Thanks !
I didn't know it was due to a smaller dynamic range. I'm a little less dumb now :D.
mikeX
28th April 2004, 14:57
little_buddha,
If you find that less or no b-frames help, you could set zones with reduced b-vop sensitivity on those areas.
A value of -35 should get rid of most b-frames
lordadmira
30th April 2004, 03:04
Or alternatively use full quality B frames. U can get pretty spectacular results with B frames set to say 8/1/0 and threshold 80.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.