View Full Version : 3ivx or xvid?
IgorC
22nd March 2005, 02:04
We have different eyes , OSPNR and SSIM too. So there is no scence to say what is better.
Sagittaire
22nd March 2005, 02:10
1) OPSNR and SSIM are objective test ... higher is better ... it's very simple.
2) H263 is best matrix for metric and by far.
3) H263 is always in the best matrix in the visual blind test from Soulhunter.
no possible discution with that ... it's reality ... :D
IgorC
22nd March 2005, 02:14
are you trying to say that objective quality is the same that visual quality?
aeternitas
22nd March 2005, 02:49
H263 matrix is best for anime encodes at bitrates designed to 1. Fit on the popular optical medium and 2. Be big enough to retain most of the detail while being small enough to not be a bother downloading. The vast majority of anime is viewed in mpeg2 and then in xvid with h263 in fansub.
If you are doing a backup, there is no reason to recompress the data, unles you like seeing what a codec can do, like me.
Maximillius
22nd March 2005, 02:58
Originally posted by Sagittaire
lol ... you don't encode complete movie but only "low motion" 50 frames: I repeat but 43 dB for Q6 encoding is typical "low motion scene"
What are you talking about? I encode complete movies all the time so I'm not sure where you are pulling all these false assumptions about me from. The point of that small test encode was to show off the fact that divx oversmoothes the picture far more then I like. Nothing more. It wasn't meant to be a comprehensive quality test. The very fact that it oversmoothes in such a low motion scene just belies how much it oversmoothes when it comes to high action scenes in the movie.
Originally posted by Sagittaire
Yes, H263 is always in the first visually (see Soulhunter visual comparison matrix) and it's probaly with PP0. With PP4 H263 is IMO and by far the best ...
Yet it doesn't mean that everyone agrees with that. Those tests are nothing more then opinions. Opinions are not facts, no matter how much you try to say they are. Yes in your opinion H263 is best, in other peoples opinions it's not.
Originally posted by Sagittaire
1) OPSNR and SSIM are objective test ... higher is better ... it's very simple.
You are joking right? Even the creators of such metrics admit that they are not always right and in many cases are wrong from perceived visual quality. There is no true objective measure of video quality as it is completely subjective. Metrics attempt to make objective measurements of video quality but they are not always right and are many times completely wrong.
IgorC
22nd March 2005, 03:45
A little test. Bitrate 700 kbit/s. DCT MMX-SSE2 , Fast compress in VD.
Divx sample :http://rapidshare.de/files/957333/3divx.mkv.html
Xvid sampls : http://rapidshare.de/files/957298/2xvid_875.mkv.html
SOURCE : http://rapidshare.de/files/957651/hp3c.vob.html
difference between sizes of sample is only 6 kbytes - 0.188% . It says
if Xvid will have 700 Mb (1CD-rip), then Divx will have 701.32 Mb
Divx6 Fusion - 3passes PSY off H263+ Insane Quality Multi B-frames + other settings by default. Divx decoder PP4 (not deringing)
Xvid 1.1 beta 1 - 2passes H.263 AQ QP GMC B-vops default chroma Opt. VHQ 4 , motion seach - 6 , VHQ for B-frames, chroma motion , trellis
best PP2 Xvid decoder (not deringing).
As you can see I donīt use special settings for Divx and Xvid like 1pass in zone Q6 , bitrate modulation, since not all people know how to use it.
Source http://x2.putfile.com/3/7920443871.jpg
Xvid 1.1 beta 1 http://x2.putfile.com/3/7920464877.jpg
Divx Fusion (last beta) http://x2.putfile.com/3/7920472530.jpg
bond
22nd March 2005, 10:34
Originally posted by Sagittaire
3) H263 is always in the best matrix in the visual blind test from Soulhunter.
no possible discution with that ... it's reality ... :Dyou will never learn that quality is subjective and ultimate "this is always best..." statements are nonsense :rolleyes:
IgorC
22nd March 2005, 20:26
Well, Sagitarieīs test isnīt so bad at all. He compared Xvid and Divx as for divx-devices compability (Xvid without qpel, gmc, packed bitstream , standard matrix H263). This way the differnce between Xvid and Divx is a smaller. But Xvid is still a bit better for me. if PSY and GMC for Divx will be fixed and some speed optimization it will be best codec ONLY for Divx devices.
Sagittaire
23rd March 2005, 00:26
In fact in my metric test XviD use the best (best for metric) setting for this HPII trailer:
- Qpel decrease metric for this trailer
- GMC is neutral (not less and not more)
- Psy is good for SSIM but not for OPSNR
In fact I use always the best setting for all codec. IMO it's very difficult to make better with XviD (IMO RC tweak are optilmal) but if XviD fan want to try ... :D
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.