DivX6 quality encoding ?

[kesako]

Hi,
A lot of people find the quality of the DivX6 codec better than older version and XviD...It's not my opinion...I think it's the worse quality of the actual codecs (H264 ateme,x264, Xvid,RV10,VP7).I've done several tries with these codecs, and DivX6 looks like bad RV9 encoding, with a smooth picture , even in high bitrate (~1200 kbps tested).
XviD let me the best impression, more than x264 : more sharp and less blocking (HVS best picture's matrix for XviD), and sometimes compatible with hardware.I haven't tested x264 on cartoons and CLI version with matrix...
At this time, XviD seems to be the best choice...
For DivX6 the only new stuff is the .divx container, perharps we have to wait for the DivX10 for quality improvements ?
Perharps someone think DivX6 is good... let show us some trailers or pictures...

[Sagittaire]

At this time, XviD seems to be the best choice...

XviD is better than x264, NDAVC and VP7 ... lol

donwlaoad this sample and see yourself
http://forum.doom9.org/showthread.php?t=90784

AVC or VP7 are very more powerfull than MPEG ASP and particulary XviD. With these sample XviD must be at 750 Kbps for the same quality than AVC at 450 Kbps ... :rolleyes:

DivX6 is always more powerfull than XviD for metric ... and visually better for the majority I think ... :rolleyes:

Actually DivX6 is the best MPEG4 ASP codec for metric ... :o

[kesako]

(~1200 kbps tested).At this bitrate , XviD is better than divx6.It's the choice of my eyes, i fuck the metrics....
At 1200 kbps , XviD can be as good as x264, perharps better with optimisations.with your samples, XviD is bad , it's ok, but for high bitrate, XviD is the most pleasing, more sharp , less block than DivX6.
Sagittaire , we don't speak about the same thing...metrics , it's a obssession for you...Are you blind ? :cool:

[Sagittaire]

Are you blind ?

Certainely you ... if you say "MPEG4 ASP is better than MPEG4 AVC" ... my eyes are perhabs very strange too ... lol

So perhabs that MPEG2 is finally better than MPEG4 ... :rolleyes:


At 1200 kbps , XviD can be as good as x264

do not mean anything ... :rolleyes:

resolution used ? average quant level used ? matrix used ?

[IgorC]

(~1200 kbps tested).At this bitrate , XviD is better than divx6.It's the choice of my eyes, i fuck the metrics....


Same here.
Metrics aren't so bad. But sometimes I also want to do that :cool:
Especially when psy 3, deblcoking 0, h263 for Xvid quant get higher SSIM and OPSNR . indeed it's only bluring and smoothing the video.

You know some guys know how to increase SSIM, OPSNR just adding some smooth and filters to codec. Why don't waste their time to increse metrics, menu, subtitles and other stuff and do something really good in quality terms?
Sagi, I thought you read about this stuff in some ebooks about video , didn't you? It's a big problem for metrics. If metrics will be ideal why devs just through their codec to SSIM OPSNR automatic researchment and 'voila' here is the best codec.

[neo_anderson]

Right Now, Nero Digital AVC is the best MPEG-4 Codec Out There! and as for asp, i truly believe my eyes, they tell me divx 6, when properly encoded (use extreme quality or insane quality mode with h.264 optimized) is definitle better and more user friendly than XVID!

[Sagittaire]

In fact XviD and DivX6 are very close for metric but DivX6 and visually too (DivX6 is always little better for metric).

IMO only Open Source Fan say "XviD is better than DivX6" but I can make real blind test (with 4CC = MP4V for all codec) and nobody could say sample A is really better than sample B or sample A is XviD / DivX ... lol

IMO at this time nobody can say "XviD is visually really better than DivX6" or "DivX6 is visually really better than XviD". Only metric can say that at this time ...

but there is a thing which is not debatable : NDAVC, x264 and VP7 are always very very better than all MPEG4 ASP codec ... :rolleyes:

[IgorC]

It's not about opens source at all. YOu may downlaod divx6 free for 6 month trial. But I don't need it. Because it's worse than Xvid to my eyes.

Sagitarries, are you sure than metrics can say you what psy, deblcoking are better? You said me that I needcompare myself differnet psy modes, cause metrics don´t enable to do that

@NeoAnderson Divx is still ASP and don't has H.264 optimized, maybe H.263+ smooth quant.

[temporance]

IMO only Open Source Fan say "XviD is better than DivX6" but I can make real blind test (with 4CC = MP4V for all codec) and nobody could say sample A is really better than sample B or sample A is XviD / DivX ... lol

4CC = MP4V for all codecs is not a complete test of the codecs. The postprocessing is also different. IMHO, DivX 6 deblocking is better (for metrics and eyes) than xvid's.

[xDrJx]

I dont contibute in which codec is the better :scared:
(I know 4 me :D )

...i truly believe my eyes, they tell me divx 6, when properly encoded (use extreme quality or insane quality mode with h.264 optimized) is definitle better and more user friendly than XVID!

To choose from 4 predefined settings may be "user friendly" but is also very limiting. And the gained quality with optimized matix and extreme/insane setting doesn't justify the loss in speed (i mean 5 fps, wtf).

[Sagittaire]

Sagitarries, are you sure than metrics can say you what psy, deblcoking are better? You said me that I needcompare myself differnet psy modes, cause metrics don´t enable to do that

[Sagittaire]

Sagitarries, are you sure than metrics can say you what psy, deblcoking are better? You said me that I needcompare myself differnet psy modes, cause metrics don´t enable to do that

1) Deblocking 0 (+/- 2 is very close for metric) is always better for metric and for my eyes (only anime exception with strong inloop 2~4 but it's particular). In blocking vs blur fight I prefer blur ... and metric too

2) psy decrease only OPSNR and increase generaly SSIM. Don't forget that psy model are too HVS mathematical model (dark masking, lumi masking) ... it's always possible to use the same model for metric evaluation like SSIM for example ... why not ... ???

3) if you want seriousely and visually compare mpeg4 asp codec make blind test with 4CC = mp4v for all mpeg4 asp codec (with mp4v for all codec it's impossible to identifie compressor used) : final conclusion

[IgorC]

I've already done a blind version. I almost always can spot difference between Xvid and Divx. As the result, Xvid is better.
Well, my congrat to you that visually deblocking 0 is better for you and for metrics. But most people prefer deblcoking -2 and even by default Ateme H.264 has deblock -2 (devs choosed it). You're a exception.

[Sagittaire]

But most people prefer deblcoking -2 and even by default Ateme H.264 has deblock -2 (devs choosed it). You're a exception.

for x264 default is 0,0
-f, --filter <alpha:beta> Loop filter AlphaC0 and Beta parameters [0:0]

and x264 dev too ... lol

but default setting do not mean anything ... VHQ1 is default setting for XviD !!!

[bond]

guys, lemme repeat the old phrase: there is no ultimatively "best" codec!

its all subjective and also not only the visual quality counts (but also cost, usability, speed...)!
eg whats the sense of using avc if you cant play it in realtime? in such a situation it indeed can be better to use xvid and not nerodigital avc...

there are simply too many factors influencing a codec, that we have found a long time ago on doom9 the compromise that everyone should make his own judgement

so guys cool down ;)

[IgorC]

I noiced the difference between the videos best for my eyes and best for the metrics is around 0.25 - 0.5 SSIM(overall 0.35) for 700-900 kbps . I think it's relative interval. It depends on individual psycovisual preference and what tecs you have - TV, HDTV, monitor 15` - 19`, and there're much other factors as source,software etc.
It should be considered this difference +/- 0.35 SSIM .That's why if 2 codecs have a diference 0.7(2* +/- 0.35) SSIM maybe some people still will prefer codec will lower metrics. And that's correct. When somebody says best for metrics the question is ' How much better '. But if the difference is more than 0.7 SSIM you can say with sure that one codec is better than other.
I just formed TOP of video codec for me considering if the difference will be less than 0.7 then codecs situated in interval of individual preferences.

SSIM
1. Ateme 81.54
2. x264 / VP7 80.50 / 80.02
3. Divx6/ Xvid/ VP6/ RV10 77.99 / 77.96 / 78.41 /77.72
4. WMV9 76.73
5. Divx3 74.40

As you can see the difference between Divx and Xvid is 0.03 SSIM. I can't spot the difference less than 0.15 - 0.20 SSIM. So 0.03 is 5-7 times less than what human eye can notice.

It's explain all discussions about quality like Xvid vs Divx, Xvid vs VP6 , RV10 vs VP6.