DivX 5.2 vs XviD 1.0.1

[Enigmax]

Look:
DivX 5.2 avi Sample Quantization H.263:
http://enigmax.en.eresmas.com/Test_DivX52.avi
DivX 5.2 avi Sample Quantization MPEG
http://enigmax.en.eresmas.com/Test_DivX52_MPEG.avi
XviD 1.0.1 avi Sample:
http://enigmax.en.eresmas.com/Test_XviD101.avi

Same script AVS. 700 kbps in both cases.

DivX 5.02 codec default Standard whit multiple B-Frames, GMC and QPEL. Quantization H.263 / MPEG-2.

XviD 1.0.1 codec default whit 2 B-Frames, QPel, GMC, VHQ 4, Trellis and Quantization MPEG.



Two pictures from the same samples from low bitrate scenes captured whit virtualDubMod (in raw video mode, no decoder and no postprocessing):
DivX 5.2
http://enigmax.eresmas.com/DivX.jpg

XviD 1.0.1
http://enigmax.en.eresmas.com/XviD.jpg

The difference is evident.

NOTE: Desactivating all postprocessing settings to compare the videos.

Greetings

[DigitAl56K]

It looks to me as if either post processing was still on for XVID, or these frames had very different quantizers.

Did you compare the frame quantizers?

[Enigmax]

No, no postprocessing. Raw video captured in VirtualDubMod.

Look these pictures captured with BSPlayer and FFDSHOW, and no postprocessing:

DivX 5.2 H.263
http://enigmax.en.eresmas.com/download/DivX_H.263.jpg

DivX 5.2 MPEG-2
http://enigmax.en.eresmas.com/download/DivX_MPEG.jpg

XviD 1.01
http://enigmax.en.eresmas.com/download/XviD.jpg

Note from the same frame the different type and the different quantizer.
XviD looks better.


Greetings

[DigitAl56K]

Thanks Enigmax,

We're currently working on quality improvements, this will help! :)

I presume this was a 007 movie, which one in particular?

[Enigmax]

Yes, "Tomorrow Never Dies".
I tried for test chapter 2.

Greetings

[nicco]

My test

Source: DV-video frameserved from premiere in yuy2
Avisynth:

AVISource("D:\prov.avi",pixel_type="yuy2")
TomsMoComp(0,15,1)
Convolution3D(Preset="movieLQ")


Divx 5.2:@900kbps
first pass

-bvn1 900 -pre 0.00,0.00,0 -psy 0 -key 300 -log "D:\prov.log" -w -b 1 -sc 50 -pq 5 -nf

second pass in slow mode

H.263- frame 407 B q:11 bits10067

http://img36.exs.cx/img36/7826/divx_1.jpg

finalsize:2.178Kb

Mpeg2 - frame 407 B q:9 bits13058

http://img36.exs.cx/img36/9241/divx_2.jpg

finalsize:2.290Kb

Xvid:
I used the setting in doom9 xvid guide http://www.doom9.org/xvid.htm at 900Kbps, H.263
except for qpel (off), max cons. BVOP's=1,and low scene impr=2%

frame 408 b q:11 (654b) (frame 407 is a P frame, but they are almost the same)

http://img36.exs.cx/img36/2325/Xvid1.jpg

finalsize:2.142Kb

Post processing disabled for both
I captured the frame in Vdub, they are jpeg images at maximum quality

[trekminal]

To Enigmax:
Please please...make these comparations without GMC and QPEL (both standalone Divx players compatible). And be less malicious choosing the picture bottom colour. :p

[nicco]

DigitAL56k:
We're currently working on quality improvements, this will help!

I think you really should work on quality improvements, in fact I can't see any noticeable quality improvements in 5.2 from 5.1.1.
You did a huge improvement from 5.0 to 5.1 so I thought it was the same for 5.1 to 5.2 but I'm a bit surprised it's not so...

[Sharktooth]

@DigitAl56K

http://forum.doom9.org/showthread.php?s=&threadid=79091&perpage=20&pagenumber=3#post524549

and

http://forum.doom9.org/showthread.php?s=&threadid=79091&perpage=20&pagenumber=3#post525031

[Enigmax]

Hello Fernando "BigFoot" (:cool: ), download the videos for an exact comparison. Nothing is malicious in then.

I always use the ASP to get maximum quality (always use HTPC), compatible with the chips Mediatek.

Greetings

[nicco]

I updated my test above;)

[Enigmax]

Capture of Sharktooht's video:

http://enigmax.en.eresmas.com/Shark2.jpg
http://enigmax.en.eresmas.com/Shark.jpg

http://valborg.hn.org/sharktooth/divx52-2000kbps-mpeg-slow.avi
The macroblocks and the artefacts are terrible, in all the sequence.

Greetings

[trekminal]

Originally posted by Enigmax
Hello Fernando "BigFoot" (:cool: ), download the videos for an exact comparison. Nothing is malicious in then.

I've got them. The difference is out there (i do not discuss this), but you are choosed the most bad image on divx sequence to distort the result, in your own way. :cool:

[Sharktooth]

I hope they'll fix it in few days or ahead's Recode will sell much more...

EDIT:
@trekminal: download my clips then...

[nicco]

Yes, it's terrible!

[Enigmax]

@Trekminal:
I've got them. The difference is out there (i do not discuss this), but you are choosed the most bad image on divx sequence to distort the result, in your own way.

No.
Then, why tried XviD with B-Frame and bigger Quantization?
The image captured is at random.
The difference is noticeable in any freeze-frame.

Greetings