VC-1 and H264

[DeathTheSheep]

Originally Posted by zambelli

Doesn't that effectively take AVC down to roughly ASP level in terms of efficiency? If that's the case, one would assume VC-1 would then have a clear edge. It'd be interesting to see a real test though.

I'd assume so at first, too, but...
There are claims by some members of this board (Sagittaire) that x264 at some of its lowest or fastest quality settings decisively outperforms XviD at its slowest and best.
You're right, it would be very interesting to see if these claims hold, and if Baseline AVC (the stuff I was talking about) is as good as max-quality XviD.

Then, if the results validate that hypothesis, we can set up a test against forms of VC-1.

[CruNcher]

DeathTheSheep in terms of Detail Preservation i can confirm this allready with the early on implementations like RV9/10 and later x264 it really preserves more details even at lowest settings and that is because of the much better implemented Qpel and partitions (not talking about 8x8 transform (FreXt) wich takes it another step above).
For x264 subme 1 allready outperforms ASP in terms of detail preservation for example XviD with vhq4 (you can enhance Detail Preservation see my EDP build but it still doesn't gets against H.264) and qpel can't stand against it you have much finer texture preservation best visible on skin textures.
and then you just have to add the facts that H.264 has no moving walls effect and ringing is almost non existant for low bitrate and then you have a winner (for low bitrate) it wont change for High Bitrate much at least the Detail Preservation wouldn't.

[zambelli]

Cruncher, I always feel like I'm going to run out of breath when reading your posts.

Originally Posted by CruNcher

ON2 allways fought against multi billion dollar companies (and Open Sourced VP3 for Theora) with less resources and does great if you ask me (just look @ the flash deal) it's research is top and now look @ the size of Microsoft Research and Compare it with On2 hehe

Compare what to what? What is the size of On2's codec team? What is the size of Microsoft's codec team? I know the answer to that second question - but I'm curious if you do, since you seem to be comparing one against the other. Not sure how though.

FYI, while MS Research has indeed contributed ideas to Microsoft codec development over the years, vast majority of the audio and video codec development is done by the codec product team in Redmond, WA. The same team also develops audio and video DSPs such as the ones found in Vista and WMP. While it's easy to think of Microsoft as one bottomless pool of people and cash, one has to understand that individual product teams within Microsoft are always finite in their resources.

everything MS does is copying all the way and that badly look for example @ Apples trailer Page and then Compare with Microsofts *rofl*

Point taken. That's just a classic case of MS being unable to get its PR act together.

But Sometimes i really wonder if MS Research is really that great for example the thing with the Xbox 360 and the Dashboard AVC playback thing lol i can only lough about this (why to buy this high tech shit if i can build me a unrestricted player) it's so ridicoulus but that's how Microsoft is restricting restricting and restricting but hey Sony isn't any better :P

Not sure why you're making a connection between MS Research and Xbox Dashboard. One has nothing to do with the other. What is your complaint about Dashboard AVC playback anyway?

[Dark Shikari]

Originally Posted by CruNcher

Eh i don't find it ok sure VP3 isn't better as VC-1 but @ the time it was released it was a match for Mpeg-4 ASP

Do people actually know when VC-1 was officially released as a completed standard?

(Hint: Long after H.264...)

[Golgot13]

Originally Posted by Dark Shikari

(Hint: Long after H.264...)

The license and patent price for VC1 was available and (exactly) same
few month after H264 (4 months after the H264 final version of patent/license price...)



Golgot13

[CruNcher]

@Dark Shikari
yes i know that but that wasn't my point my poin't was that VP7 would have been existed anyways but you compare VP3 with VC-1 that's strange and even if VC-1 wouldn't exist VP7 would as it is a direct contender to H.264 and VC-1 is just Microsofts specialy modified version of H.264 (with research they hold back in the H.264 dev process) so the point is without VC-1 we wouldn't stuck with VP3 (as you said H.264 was long before VC-1)

@zambelli

Compare what to what? What is the size of On2's codec team? What is the size of Microsoft's codec team? I know the answer to that second question - but I'm curious if you do, since you seem to be comparing one against the other. Not sure how though.

you right i don't know how big the Windows Media Video codec dev team of Microsoft is but for sure the Mother company has more resources to spend then On2
but that's also not the point the point is how smart are the guys in each of both teams :P very good example is CoreAVC (1 main dev beets the whole R&D of Elecard/Mainconcept Cyberlink/Intervideo and even races Nvidia and Atis First Generation Hardware acelleration away and that in under 2 weeks of development i find this quiet funny don't you ).
And what does Microsoft to compensate such things and that since years they buy other innovative companies with such brains in them like (Seadragon,iview)

Not sure why you're making a connection between MS Research and Xbox Dashboard. One has nothing to do with the other. What is your complaint about Dashboard AVC playback anyway?

Because i estimate with such a big R&D behind your backs
it would be no problem to code a very efficient AVC Dashboard Decoder that can make use of the full power of the Xbox 360 multicores (a system Microsoft even designed) to playback 1080p @ higher bitrates then 15 Mbps or 10 what you say in your faq is the maximum and this 4 GB limit (lol)

AVC:

A: Video Bitrate: 10 Mbps with resolutions of 1920 x 1080 at 30fps. See the question about max bit rate, resolution, and frames per second.

AVC/ASP/SP
A: The maximum file size for an MPEG-4 Part 2 and H.264 file that can be played is 4 GB. However, Windows Media Player 11 and the Zune software do support streaming WMV files greater than 4 GB.

Those restrictions sound to me like you haven't your own Design under Controll and are sorry to say (blind) to improve it
it really sounds ridicoulus for what the Power this Machine has and o wonder is able to playback HD-DVD bitrates 30 Mbps from the HD-DVD Player or are these values only for DVD-Drive playback (this i could understand) but nothing is mentioned of HDD or DVD drive in this faq so @ the moment Sonys PS3 seemes less restricted in Video Playback Power then the Xbox 360 to me sad sad

Why do you think people create firmware hacks or try to run Linux, yes lifting such restrictions is the answer, because the Power of these machines allows more and to see that the Vendor that creates them doesn't make fully use of them is hurting some people (you can call them Home Researchers ).

Ah and yes

VC-1

A: Video Bitrate: 15 Mbps with resolutions of 1920 x 1080 at 30fps. See the question about max bit rate, resolution, and frames per second.

Big marketing gag hey look VC-1 can be playbacked with 5 Mbits more then AVC it's less Decoding Complex on our *rofl* highend (Mobile) Decoding Machine o come on this is so ridicoulus

[Dark Shikari]

Originally Posted by Golgot13

The license and patent price for VC1 was available and (exactly) same
few month after H264 (4 months after the H264 final version of patent/license price...)



Golgot13

But VC-1 was designed as a competitor to MPEG-4 ASP, not AVC... it came a little late did it not?

[benwaggoner]

Originally Posted by Dark Shikari

But VC-1 was designed as a competitor to MPEG-4 ASP, not AVC... it came a little late did it not?

It was designed to be a good codec, not to beat any particular existing standard .

You can think of the design goal of VC-1 to be the best possible compression efficiency, including for HD scenarios, with appropriate performance to both PC and ASIC decoding.

[benwaggoner]

Originally Posted by Dark Shikari

Agreed. Personally I think VC-1 is useless; its only advantage is lower playback processor requirement, which it gets at the cost of a massive loss in quality (or better said, a much higher bitrate required for the same quality).

Okay, challenge accepted . Make a "massively" better clip with the same parameters as the 2 Mbps 720p clip on my blog:

http://on10.net/Blogs/benwagg/elepha...-720p--2-mbps/

Let me know when your comparative clip is avaliable.

[Golgot13]

Originally Posted by benwaggoner

Let me know when your comparative clip is avaliable.

I hope that Manao or Sagitaire can make a good "encoding" answer.
And I hope that the Ateme team (good developper) can answer too by a nice encoding file...





Golgot13

[vsv]

Originally Posted by benwaggoner

Okay, challenge accepted . Make a "massively" better clip with the same parameters as the 2 Mbps 720p clip on my blog:

http://on10.net/Blogs/benwagg/elepha...-720p--2-mbps/

Let me know when your comparative clip is avaliable.

Ben, you can find on this forum trailer of Batman Begins 1920x1080(816) encoded two years ago in h264 (ateme) at 3Mbps.
Elephant Dreams coded by x264 720p24 2Mbps may be better than vc1

[benwaggoner]

Originally Posted by Golgot13

I hope that Manao or Sagitaire can make a good "encoding" answer.
And I hope that the Ateme team (good developper) can answer too by a nice encoding file...

Okay. Of course, the longer it takes, the better our encoder will get in the interim...

The version everyone is working with is over a year old at this point, and there's no reason to expect we aren't advancing as fast as x264 .

[jmnk]

Originally Posted by vsv

Ben, you can find on this forum trailer of Batman Begins 1920x1080(816) encoded two years ago in h264 (ateme) at 3Mbps.
Elephant Dreams coded by x264 720p24 2Mbps may be better than vc1

Would you happen to have a valid link to this encode? I found one from 2005 here http://forum.doom9.org/showthread.ph...787#post656787 (which I think it is what you are referring to) - but the link seems to be no longer valid?

[DeathTheSheep]

The version everyone is working with is over a year old at this point, and there's no reason to expect we aren't advancing as fast as x264

Mind providing an updated encoder (preferably simple exe) so we can actually test it against the "new guys?" Thanks in advance.

[Sagittaire]

Originally Posted by benwaggoner

Okay, challenge accepted . Make a "massively" better clip with the same parameters as the 2 Mbps 720p clip on my blog:

http://on10.net/Blogs/benwagg/elepha...-720p--2-mbps/

Let me know when your comparative clip is avaliable.

Well I can make that quickly with x264 at 720p. But you don't use the best possible VC1 implementation here.

- Your encoding have major blocking problem in high motion part.
- Your encoding have major problem with fade scene because WMV9 AP don't use adaptative bframe and bframe in fade scene produce bad result.

I think that VC1 can produce very better result at 2 Mbps.

[Tack]

Originally Posted by benwaggoner

Let me know when your comparative clip is avaliable.

From which version of Elephant's Dream did you create your encode? So that people are working from the same source.

Also, aside from peak bitrate which has some rationale, why should the same settings (e.g. number of B-frames) be used?

[Sagittaire]

From which version of Elephant's Dream did you create your encode? So that people are working from the same source.

Source is the 1080p RGB uncompressed Master with YV12 (rec 709?) color matrix convertion

Also, aside from peak bitrate which has some rationale, why should the same settings (e.g. number of B-frames) be used?

Average bitrate at 1863 Kbps for video. Peak bitrate at 8000 Kbps with large buffer. Well I think that here Rate Control is unrestricted with these parameters.


Here the AVC/AAC/MP4 encoding
1280x720x24 AVC at 1863 Kbps
5.1, 48 Khz, 16 bits AAC at 128 Kbps

For example no problem for the first scene (fade scene). Less blocking for high motion. More Sharp too I think. It's really quick encoding with x264 with classic profil and without particular optimisation.

[Tack]

Originally Posted by Sagittaire

Source is the 1080p RGB uncompressed Master with YV12 (rec 709?) color matrix convertion

Where can I download that? The only thing I've been able to find is 22G worth of individual png files. Which would clearly be the highest quality source to work from, but 22G is a large chunk of my monthly bandwidth quota to use up.

... looking at the "HD 1080p24 Challenge" thread I see this is indeed the source you worked from.

[Sagittaire]

Originally Posted by Tack

Where can I download that? The only thing I've been able to find is 22G worth of individual png files. Which would clearly be the highest quality source to work from, but 22G is a large chunk of my monthly bandwidth quota to use up.

PNG source is the good source ...

[benwaggoner]

Originally Posted by Sagittaire

Well I can make that quickly with x264 at 720p. But you don't use the best possible VC1 implementation here.

- Your encoding have major blocking problem in high motion part.
- Your encoding have major problem with fade scene because WMV9 AP don't use adaptative bframe and bframe in fade scene produce bad result.

I think that VC1 can produce very better result at 2 Mbps.

Yeah, I just used the released codec since it was mainly as a tutorial about using the registry keys with it.

Assuming some good H.264 versions pop up, I'll go back and do a new one with the new codec.