DIVX 5.1 and "Q"

[Candock]

Gej just was on german television :o

i like the idea of going away from the iso-mpeg4!
it's good if they get higher quality and if it's a good way to achieve that
why not?
and it's also playable

[SeeMoreDigital]

Originally posted by Candock
Gej just was on german television :o
i like the idea of going away from the iso-mpeg4!
it's good if they get higher quality and if it's a good way to achieve that why not?
and it's also playable What! Is this what Gej has said. Are DivX really going to dump the ISO Mpeg4 standard? If so, did Gej mention what they will be using. Is it even Mpeg4?

Cheers

[Candock]

NO, he hasn't said that!!

i said, that it would be good :).
the future is not mpeg4 (and of course not xvid :p) anymore.
and i would love to see a cross-standard of different codecs.
something like kludge or rududu!

[BoNz1]

Originally posted by Candock
NO, he hasn't said that!!

i said, that it would be good :).
the future is not mpeg4 (and of course not xvid :p) anymore.
and i would love to see a cross-standard of different codecs.
something like kludge or rududu!

I do not agree, sure mpeg4 ASP might be beginning to show its age. But current implementations such as XviD are more than competitive with any other codec today. You are right maybe in a year or two mpeg 4 ASP will no longer be the best choice. However, to suggest we completely abandon standards is ridiculous. Standards are extremely important and I think DXN recognizes this. In order for a video codec to be truely successful today it is important there is a degree of interoperability. I sincerely hope that divx6 will be mpeg4 AVC (h.264) and not some mismatch codec that might only play on certain platforms and environments.

[Prettz]

If they drop from the MPEG4 standard will this mean that there will no longer be backward compatibility? If people aren't able to play all their old divx files then I think we can rest assured that DXN will get absolutely zero business from normal consumers. They may be able to sell hardware to corporations who don't know any better, but that will probably be it.

[Crabba]

@Sagittaire: You really need to understand that blurring the AVS source and then doing a PSNR test that shows lower PSNR has nothing whatsoever to do with if an encode of that AVS source will get better or worse PSNR. Of course the blurred source will get worse PSNR than the unblurred source, but if you encode both the unblurred and the blurred avs-scripts you will probably see that the blurred encode will get a higher PSNR when compared to the blurred AVS script than the unblurred encode compared to the unblurred AVS script.

I have done tests with this but not with blur but using different resizing methods instead and not surprisingly I got higher PSNR when using bicubic resize with 0.5 than when using Lanczos. I encoded a 50 minute program and got about 0.2 better PSNR when using bicubic 0.5 compared to Lanczos. And that's just when using two quite sharp resizing methods, the difference would of course be a lot more when blurring etc.


@All:

I did a full test encode with DivX 5.1 as well tonight (best possible quality) and did a PSNR test this morning which was quite disappointing:

Total frames processed: 70950

DivX 5.1:
Minimum Average Maximum
Mean Absolute Deviation: 0.0182 1.5912 4.0431
Mean Deviation: -0.4888 +0.1023 +0.7270
PSNR: 32.8678 41.1957 65.2675

XviD Koepi 0624:
Minimum Average Maximum
Mean Absolute Deviation: 0.0082 1.4472 3.5595
Mean Deviation: -0.4691 +0.0023 +0.3666
PSNR: 33.3148 42.0160 68.9328

WMV9 VCM:
Minimum Average Maximum
Mean Absolute Deviation: 0.0145 1.3855 3.7924
Mean Deviation: -0.2991 +0.0104 +0.3366
PSNR: 32.9762 42.6142 66.4800

It's also visually inferior to both XviD and WMV9...

After comparing frames and also watching through the entire video of all three codecs I think I'm going with WMV9 after all.

[temporance]

> I did a full test encode with DivX 5.1 as well tonight (best possible quality) and did a PSNR test this morning which was quite disappointing:

Uh? My PSNR tests between 5.1 and WMV9 are not consistent. WMV9 seems to have the lead only at lowest bitrates (e.g. 350kbit/s when both codecs look pretty bad). DivX 5.1 (and xvid for that matter) beats WMV9 at the highest bitrates in terms of PSNR (and I'd trust PSNR more when the artifacts are smaller).

In terms of visual appearance (no postprocessing), there's no comparison - 5.1 rocks at all bitrates!! (IMHO of course ;) )

What do you mean by "best possible quality" - and how doe you get these settings? Is that best possible settings for PSNR, or best possible settings for visual appearance. E.g. psychovisual (and sometimes preprocessing) will produce lower PSNR but output will look much better.

What bitrate did you use and what was the nature of your source?

[Sagittaire]

Post Processing of WMV9 must be desactived and WMV9 is better than iso-MPEG4 codec ... for all bitrate in PSNR, SSIM, VQM and visual test ...

iso-MPEG4 codec (XviD and DivX) have specifications who limited performance. RV9 and WMV9 are better than iso-MPEG4 codec ...

Encoding Two Towers with RV9 and WMV9 is possible (with no blocking and no ringing) on one 80 min cdr ... try with iso-MPEG4 ... ;-)

[temporance]

Originally posted by Sagittaire
WMV9 is better than iso-MPEG4 codec ... for all bitrate in PSNR, SSIM, VQM and visual test ...Sagittaire: you really must learn to say "IMHO" when you make that sort of claim. IMHO. ;)

[Sagittaire]

Sorry ...

WMV9 is better than iso-MPEG4 in PSNR, SSIM, VQM test (mathematical test thus not IMHO)

WMV9 is better than iso-MPEG4 for visual test, IMHO of course ... ;-)

[Lefungus]

It's almost good Sagittaire.
Now add that WMV9 is better in mathematical tests, not generally but on a limited set of videos. :)

[SeeMoreDigital]

Originally posted by Sagittaire
... WMV9 is better than iso-MPEG4 for visual test, IMHO of course ... Guy's. I can't help feeling that we are all 'going over old ground here'. As I think it should be a well established fact now that Mpeg4 DivX/XviD really is not as good as WMV9 at bitrates below 950kbps (approx). When encoding above 950kbps, DivX begins to respond better than WMV9. This carries on right upto around 1450-1500kbps. After this it's fairly even.... In My Humble Opinion!

Seriously though. I am hoping that DivX 5.1 will offer greater low bitrate encode improvement over 5.0.5. As it would be nice to see DivX encodes looking better than WMV9 at around 850kbps, instead of 950kbps!

Cheers

[Crabba]

Ok, here is some more info on the settings in my test:

Source: Blue planet episode 1 DVD
Really tough codec test material.

Bitrate for WMV9: 1836, bitrate for XviD and DivX5.1: 1880.
WMV9 and XviD produced almost perfect filesize, while DivX5.1 (and Kaukura as well) produced a 1-2mb oversized file.
res: 656x368
length: 48:33
gknot bits/pix: 0.312
calc. video size: 636mb

AVS script:
FieldDeinterlace()
crop(14,4,692,566)
BicubicResize(656,368,0,0.5)

DivX5.1 settings: 2pass, B-frames, GMC, Q-Pel, slowest, kf 250, psychovisual: slow with 'DivxPvSimpleChroma.txt' enabled, the rest at default. (I also did the same test with Kaukura without psychovisual - slightly higher PSNR but still lower than XviD & WMV9)

XviD koepi settings: 2pass, Motion 6, MPEG, VHQ 4, kf 250, Qpel, chroma motion, 2 bframes, rest at default.

WMV9: 2pass vbr, complex, performance 100, kf 16000, rest at default.

All tests done WITHOUT any post-processing as that has nothing to do with actual video quality IMHO!

Result:

DivX5.1 - Really blocky in some high action scenes, pretty good quality & detail otherwise.

XviD - Really nice quality & detail most of the time, but chokes on some tough underwater-scenes which looks pretty ugly.

WMV9 - Really nice quality & detail most of the time, with some ugly smoothed out keyframes and some other area smoothing that isn't really noticable unless you're looking for it. But it's not really noticable when playing the video in realtime, so it's not as disturbing. Handles under-water scenes well.

[Sirber]

no RV9 test? :(

[Sagittaire]

Use this reg for max quality with WMV9
http://jfl1974.free.fr/Video/PP-WMV9.reg

PP=0 -> Post Process desactived
PP=4 -> Post Process max

The post process for WMV9 isn't a deblockink/deringing PP like MPEG4 PP in DivX decodeur or ffdshow but a smoother ... ;-)

The avi container isn't ideal container for WMV9 ... use wmv or mkv ...
http://jfl1974.free.fr/Video/Matroska-WM9.rar

Demo with:
-WMV9 715 Kbps 640*272 WMEncodeur 9
-WMA9 Pro 5.1 44.1 Khz 16 bit 132 Kbps WMEncodeur 9 french & english
-SSA Subtitles french & Infos
-Matroska Container

[SeeMoreDigital]

Originally posted by Sirber
no RV9 test? :( I hope not here my friend!

I think it's a given now that RV9 will look better! Just DivX 5.1 and Q questions and answers here please!

[temporance]

Originally posted by Crabba
DivX5.1 settings: 2pass, B-frames, GMC, Q-Pel, slowest, kf 250, psychovisual: slow with 'DivxPvSimpleChroma.txt' enabled, the rest at default. (I also did the same test with Kaukura without psychovisual - slightly higher PSNR but still lower than XviD & WMV9)
For me, Q-pel gives no advantage, especially at lower bitrates and when there are high-motion scenes. Also "fast" PV performs better than "slow". Finally, use the bitrate modulation if you're sensitive to artifacts in high motion scenes.