Log in

View Full Version : Which xvid settings in the first pass affect the result most?

kypec
19th March 2009, 11:30
I'm always doing my encodes in two passes with xvid. Now I'm wondering which settings play the role in the resulting .stats file which will be used in the second pass. I did a short test with 4 configurations where different options were enabled:
None
VAQ only CQM only "EQM_V3HR.xcm" VAQ+CQM

The statistics produced for the same clip with I/P/B frames count and (their average sizes) summarized in the log files:
29/556/915 (56860/31596/5151) [difference in %]
29/556/915 (47524/25498/5310) [-16/-19/+3]
29/566/905 (92482/38659/7249) [+63/+22/+41]
29/569/902 (82060/31747/7440) [+44/+0/+44]

I know that the main goal of CQM is to retain more details in the picture yet VAQ serves the same purpose, doesn't it?
Am I right to conclude that VAQ alone improves the compressibility most and its use is therefore not recommended with CQM together?
Sharktooth
19th March 2009, 18:36
some CQMs along with VAQ produce unwanted results.
my advice is to not use CQMs with VAQ but it seems also, some CQMs dont cause any problems. it all depends on the coefficients distribution.
so if you have some time to spend, test the CQM+VAQ combo on some clips of the source you're going to encode and see if the results are pleasing.
Soulhunter
20th March 2009, 17:21
Also kepp in mind that improving the compressibility is not necessarily the same as improving the resulting quality... ;)
BigDid
21st March 2009, 00:15
...
I know that the main goal of CQM is to retain more details in the picture yet VAQ serves the same purpose, doesn't it?
Am I right to conclude that VAQ alone improves the compressibility most and its use is therefore not recommended with CQM together?
Hi,

I agree with Sharktooth on testing needed.

I made a comparison a while ago on a 1 pass encode that was aimed at compressibility difference; see:
http://forum.doom9.org/showthread.php?p=1250506#post1250506

Although aimed to tackle the undersize/maxxing Xvid-codec notion, it gives some info on compressability for H263 matrix and the Sharktooth Eqm-v2 cqm.

I have also done some tests on the Eqm V3-HR, the HR version modded by Heini and Didée 6of9 all giving good quality and compressability results with VAQ.

Atm I use Xvid with Eqm-V2 (or H263 when needed) and Heini-HR in the 1000/1500Kbpp/s range. For above I go x264.

Did