Third Annual MSU MPEG-4 AVC/H.264 Video Codec Comparison

[heffeque]

HERE (http://compression.ru/video/codec_comparison/mpeg-4_avc_h264_2006_en.html)'s the link to the web page and HERE (http://compression.ru/video/codec_comparison/pdf/msu_mpeg_4_avc_h264_codec_comparison_2006_eng.pdf)'s the PDF of the famous Moscow University comparison.
x264 does very VERY good. It's only flaw: too much CPU required compared to it's most close opponent: MainConcept's h.264 implementation.

Good job! ;)

[akupenguin]

Duplicate of http://forum.doom9.org/showthread.php?t=112686.

too much CPU required compared to it's most close opponent: MainConcept's h.264 implementation.
No, MainConcept used too little CPU. ;)
In order to try to compare the quality of different codecs at the same speed, MSU asked for the (ssim-wise) best settings within a certain quota of CPU-time (see the pdf, page 7). I gave such settings. It's not my fault if MainConcept ignored the guideline and gave a faster setting, or is not sufficiently configurable to fine-tune the speed.


Originally Posted by Alexander Parshin on 2006-8-11
We have tested speed of x264.
Results are the following:

HDTV
- high quality 0.6 fps (minimum 0.4 fps)

Movie
- high speed - 11.3 fps (minimum 10 fps)
- 1-pass high quality - 4.4 fps (minimum 4 fps)
- 2-pass high quality - 1.8 fps (minimum 1 fps)

VideoConference
- high speed - 71 fps (minimum 70 fps)
- high quality - 27 fps!!! (minimum 30 fps)

[...] The only problem is "VideoConference High Quality" preset, where you are a little slower than our requirements. Is it possible to speed up that preset a little?

Originally Posted by Loren Merritt on 2006-8-15
OK, this should be about 12% faster: [...]

[Sergey A. Sablin]

No, MainConcept used too little CPU. ;)
Yes, we do. Nevertheless with this too little CPU usage ssim-wise MainConcept was better for that time, isn't it?

In order to try to compare the quality of different codecs at the same speed, MSU asked for the (ssim-wise) best settings within a certain quota of CPU-time (see the pdf, page 7).
There was a restriction for minimum speed to encode, which doesn't mean all codecs shall be exactly same speed - just not slower ;)

I gave such settings. It's not my fault if MainConcept ignored the guideline and gave a faster setting, or is not sufficiently configurable to fine-tune the speed.
Neither it is our, cause the may be another interpretation: x264 configurable settings for quality doesn't cost the time they waste. It is pretty clear for me from the results: a bit lower quality + at least 1.5 times slower. Another opinions?

[chadamir]

You really ought not to be smug considering you've had a paid team working on this project for a considerable amount of time to get it to slightly better quality as opposed to an open source project worked on at someones leisure. Your product also lacks some of the options available in x264 and lacks a command line interface. It's also 400 dollars which is not a consumer friendly price, despite being advertised as for consumers.

All of that being said, I don't understand the point of this fighting. The MSU test lacked all the codecs they said they'd have anyway (didn't they think they were going to have like 30?) Also, I dont feel they used enough scenes to justify their results as being statistically significant, do you? Also, one has to ask if the difference in psnr is even noticable.

[heffeque]

I would have liked seeing DivX 6.5 and the latest Xvid. Also Microsoft's VC-1 just to see how lame M$'s codec is.

[Sergey A. Sablin]

You really ought not to be smug considering you've had a paid team working on this project for a considerable amount of time to get it to slightly better quality as opposed to an open source project worked on at someones leisure. Your product also lacks some of the options available in x264 and lacks a command line interface. It's also 400 dollars which is not a consumer friendly price, despite being advertised as for consumers.All of that being said, I don't understand the point of this fighting. The MSU test lacked all the codecs they said they'd have anyway (didn't they think they were going to have like 30?) Also, I dont feel they used enough scenes to justify their results as being statistically significant, do you? Also, one has to ask if the difference in psnr is even noticable.

if you didnt caught I just answer to accusations Loren did. So cool down. The only one guy who is fighting here is you.

I may say you a lot about why commercial software is payware, why CLI isn't useful for commercial usage and how it is easy to implement it and how long we have it, about competence of MSU team, about the features each codec lacks. But I won't, cause it is just wasting the time. Everybody knows these, but some guys wants to make some things looks differently.

[Sagittaire]

Your product also lacks some of the options available in x264 and lacks a command line interface. It's also 400 dollars which is not a consumer friendly price, despite being advertised as for consumers.


1) The CLI encoder form Mainconcept/Elecard exist too (like Ateme)

2) Gui is always more friendly user interface

3) The CLI version have simply more complete option than x264
- Multipass encoding (like for x264)
- Complete VBV setting (x264 work only for CBR)
- baseline, Main, High and High444 profil
- Lossless Mode
- interlacing Mode
- WeightedPred (better than x264 WeightedBiPred)
- Complete flag setting (HDDVD and BD compatibility)
- real Psy mode (better than x264 AQ)

... and IMO better speed at the same quality level

4) IMO is a good H264 codec for professional ... price is not very high if you compare with Sonic Cinevison for example.

[Revgen]

Neither it is our, cause the may be another interpretation: x264 configurable settings for quality doesn't cost the time they waste. It is pretty clear for me from the results: a bit lower quality + at least 1.5 times slower. Another opinions?

x264 is free.:D

[R3Z]

1) The CLI encoder form Mainconcept/Elecard exist too (like Ateme)

2) Gui is always more friendly user interface

3) The CLI version have simply more complete option than x264
- Multipass encoding (like for x264)
- Complete VBV setting (x264 work only for CBR)
- baseline, Main, High and High444 profil
- Lossless Mode
- interlacing Mode
- WeightedPred (better than x264 WeightedBiPred)
- Complete flag setting (HDDVD and BD compatibility)
- real Psy mode (better than x264 AQ)

... and IMO better speed at the same quality level

4) IMO is a good H264 codec for professional ... price is not very high if you compare with Sonic Cinevison for example.


Does the mainconcept encoder suffer from the dreaded blocking and banding of smoke and dark scenes like x264 ?

[dukey]

be nice to see a comparison in pictures

[Pookie]

Mainconcept h264. Only $600

Transcoding your video with X264 for free - Priceless:D

[Sagittaire]

Mainconcept h264. Only $600

Transcoding your video with X264 for free - Priceless:D

It's really not high for a professional product. You know the price for Sonic Scenarist or Sonic Cinevision ... ???

[Pookie]

Yeah, I know. Just trying to make the point that it doesn't have to be commercial to be great. ;)