[HD 1080p24 Challenge] MPEG2, VC-1 and H264 with real uncompressed source

[Sagittaire]

Oops! ;)

Since Motion JPG-2000 is used in theatres , I wonder how it got excluded from the specs for HD DVD? Another one of M$'s gambits?

http://www.groklaw.net/article.php?story=20070829070630660

No MPEG2, H264 and VC1 produce simply better result at the same bitrate. Pure intra coding is not a really efficient way for best quality/bitrate performance ...

[CruNcher]

@benwaggoner
any estimation when you gonna release the source you talked about ?

[Golgot13]

@benwaggoner
any estimation when you gonna release the source you talked about ?

Ben can not release the source now because it's period of show and seminar (IFA, IBC, CEDIA,...), we need to wait 1-2 week
I will may be see him and talk about the VC1 efficient (some people on this forum made test with movie grain and... :D :p )

[zambelli]

Oops! ;)

Since Motion JPG-2000 is used in theatres , I wonder how it got excluded from the specs for HD DVD? Another one of M$'s gambits?
You really overestimate Microsoft's influence in the HD-DVD forum and the video industry in general. For most part Microsoft is considered an underdog in the video industry and is far from commanding others what to do. The OOXML is completely unrelated to Microsoft's video efforts. Yeah, I know, same company, so clearly it's all related, right? Well, no. There's really no coordinated effort going on on that level.

[Manao]

MJPEG 2K was not considered for HD DVD because it requires a bitrate of 60+ Mb/sec. That means 50 GB just for a 2 hours long movie.

Cinema uses MJPEG 2K because it's simpler. They can afford the bitrate ( the limit is the harddrive bandwidth, not its capacity ), and they want visually lossless. At the time they took the decision, H264 was just being standardized, didn't have 8x8 DCT, was as complicated as JPEG2K to decode in intra, and was poorly rated against Mpeg2 at high bitrates.

[CruNcher]

Ben can not release the source now because it's period of show and seminar (IFA, IBC, CEDIA,...), we need to wait 1-2 week
I will may be see him and talk about the VC1 efficient (some people on this forum made test with movie grain and... :D :p )

Yes and im making steady progress i found what could be a problem in x264s block prediction heavy use of intra blocks @ low subme ;) and also other things are on my list but i have a real life too :P and also testing some brandnew HVS stuff i can't talk about yet ;)

[cannuck]

Yeah, I know, same company, so clearly it's all related,

Well that's a good guess;) It's that simple for some people (like me). Again, it's not just OOXML, if you check out GROKLAW, you will easily see that the issue(s) is much larger. There are allegations there, about M$ trying to destroy FOSS, and Linux "communities" - which (if true) would have a profound impact on everything.

I assume that you are in a "Pay For Say" situation, so you can't say anything different than what you are saying about this issue.

[zambelli]

I assume that you are in a "Pay For Say" situation, so you can't say anything different than what you are saying about this issue.
There's nothing different to be said about it. I've been working at Microsoft for 5 years and I have never heard anybody talk about destroying FOSS or Linux, especially not on a product team level. There is no dirty secret to tell because there is no dirty secret to begin with.

[Sagittaire]

Well that's a good guess;) It's that simple for some people (like me). Again, it's not just OOXML, if you check out GROKLAW, you will easily see that the issue(s) is much larger. There are allegations there, about M$ trying to destroy FOSS, and Linux "communities" - which (if true) would have a profound impact on everything.

I assume that you are in a "Pay For Say" situation, so you can't say anything different than what you are saying about this issue.

Well I don't like always the same simplist "M$ is evil" and "Open Source is beautifull". We live in a real world and in this world we must pay the factures. In this real world MS must pay its developers. In this real world MS must simply make money just like all the other company in the world.

Anyway in this challenge Open Source (x264 implementation) produce better objective result than MS (PEP implementation).

[CruNcher]

There's nothing different to be said about it. I've been working at Microsoft for 5 years and I have never heard anybody talk about destroying FOSS or Linux, especially not on a product team level. There is no dirty secret to tell because there is no dirty secret to begin with.

hehe zambelli you know this ;)



/\
/ \
/ \
/ \
/ \
/ \
/ \
/ \
/ \
/ \
/ \
/ \
/________________________\

[Golgot13]

I've been working at Microsoft for 5 years and I have never heard anybody talk about destroying FOSS or Linux, especially not on a product team level.

This thread is not to talk about this....
But last problem of MS patents use by open source like Linux...:rolleyes:
After some Linux companies signed agreement with MS.... ???

There is no dirty secret to tell because there is no dirty secret to begin with.

In all company there is some secret, it's business.


I don't see Ben at IFA... :(
I don't see any link for video source with grain to close this Summer Challenge...

[Beastie Boy]

It would be great if this thread could get back on topic and concentrate on producing some results.

Cheers, Beastie.

[Sharktooth]

Results have been produced.
Look at the updated conclusions on the first page.

[Beastie Boy]

Sorry. I thought the idea was for people to try and better the previous results and come up with some optimal settings for each bitrate / codec combination. Has this already been done? If so, I need to re-read this thread and take a look at the videos produced.
And on with the OT rantings :D

Cheers, Beastie.

[Sharktooth]

yes, you're right, but it seems no one has produced better results than the actual ones.

[Golgot13]

Sorry. I thought the idea was for people to try and better the previous results and come up with some optimal settings for each bitrate / codec combination.

The result is available but MS said it was not good because video source is CG.
I made some test:
H264 is better than VC1 from Cinevision PSE (but X264 can not produce better quality
than Mainconcept core :) ).

We wait a video source with grain from MS. It will available soon in middle of September
after all shows and seminars.

[Dark Shikari]

(but X264 can not produce better quality than Mainconcept core :) ).
Expect this to change soon :)

[Sagittaire]

Annexe - Update

05.09.2007 - New VC1 encoding
11.08.2007 - All the H264 streams are availables
05.08.2007 - Max adaptative bframe at 7 for VC1, H264 and MPEG2
04.08.2007 - Unlimited range vector for VC1
03.08.2007 - Encoding with libavcodec at 18 Mbps
02.08.2007 - Max adaptative bframe at 7 for VC1
01.08.2007 - Complete Test with x264, pep and libavcodec


I use always pep 1.06 with really insame profil this time. Only one thread, 1 bframes, high search level for first pass, highest possible level search. I use perceptual option because it's better for SSIM and Eyes too.

This profil produce really better OPSNR. I don't use PP1 for this new profil: SSIM is really worst than previous result for 6Mbps encoding. PP is really usefull for "low bitrate" (aka high quantizer encoding). PP for VC1 is not HDDVD/BD compliant.

Zambelli from MS produce stream with pep 1.07 with little better result. 12 Mbps encoding produce 46.71 dB vs 46.30 dB for OPSNR and 92.71 vs 92.25 for SSIM. Anyway pep 1.07 can certainely produce little better result because Zambelli stream don't respect the bitrate interval for this challenge ( 940 000 Ko for 1.07 vs 958 000 Ko for 1.06).

[Golgot13]

Hi all,

a post to remember at Ben we wait "the" source.
I don't have any true answer if I will can use the new core 1.07 of PEP.
I sure Ben will use it to made encoding file (but I prefer to test it too).

I made some test of VC1 on good film source (without grain ;) because digital movie)
I really surprise by the good quality of VC1 at 8Mbps on this source, but H264 version is better :p .

[CruNcher]

@Golgot13

Encoding/Decoding Complexity of both compared vs the Quality Difference ? was it far away ?

[Golgot13]

@Golgot13
Encoding/Decoding Complexity of both compared vs the Quality Difference ? was it far away ?

Only to finish this thread.
Some people are not agree to say what it is the best codec.
Ben said it is not same result with video with grain (movie grain) and promised a video source.

There is some reason that the industry (broadcast, IPTV box,...) choosed H264
May be some people can say on IBC if there were H264 solution for broadcaster TV, mobile provider,...
(If I say it some people will not believe me)

[benwaggoner]

Hi all,

a post to remember at Ben we wait "the" source.
I don't have any true answer if I will can use the new core 1.07 of PEP.
I sure Ben will use it to made encoding file (but I prefer to test it too).
I think I said I'd try to get it done "a few weeks after IBC" and IBC just ended today :).

I've got all the source on a drive, but need to build my new workstation first.

[Golgot13]

... and IBC just ended today :).



:eek: :eek:


I've got all the source on a drive, but need to build my new workstation first.

Intel Quad Core + NVidia card 8500 (at minimum) will be enough for VC1 and H264 encoding :rolleyes:
The really problem of HD encoding is the size of uncompressed video (size and speed of hard disk).
My dream for HD encoding it's a hardware HD encoder like BitPack of digital vision
which can encode in VC1, MPEG2 and H264.

[Golgot13]

I think I said I'd try to get it done "a few weeks after IBC" and IBC just ended today :).

What sort of sequence there are on this video source ?
Movie with 3D, war, paysage,.. ?

I saw last bad new for MS today in EU. I hope you will not have same problem with VC1 toolkit
given only at some specific companies 6 or 9 month before others.

[Sharktooth]

ill define it a good news. EU is not US... and the time of the US ubercorporations ruling the world without respecting any laws is coming to an end.

[Golgot13]

About VC1 vs H264, after other tests this week: H264 can not be compared with VC1.
H264 is better, last developpment for movie grain with or without FGM (or FGT) are good.
And a new actor (great actor) of H264 will propose soon a encoder for HDDVD and BD (only H264).

It will be sure better than Sony, Thomson, CC-HDe,...

About IBC, I surprise that nobody said (except me) there was lot of H264 hardware.
And H264 is used on different broadcast system: broadcast TV by terrestrial, cable, satellite;
IPTV (HD in France) ; mobile phone (all last phone support H264 not VC1 except phone with WinCE OS);
video security system; ...

[Sharktooth]

yeah, i know that. why do you tell this to me?
maybe my post was confusing... i was referring to M$...

[bobololo]

About IBC, I surprise that nobody said (except me) there was lot of H264 hardware.
And H264 is used on different broadcast system: broadcast TV by terrestrial, cable, satellite;
IPTV (HD in France) ; mobile phone (all last phone support H264 not VC1 except phone with WinCE OS);
video security system; ...

Actually that has not changed over past years. I would say that H.264 has took over VC-1 on the broadcast market for 2 years in my opinion.

I made a quick stop at Microsoft booth which was rather hidden compared to the main show floor. It was a real labyrinth to get to the topaz lounge ;) Anyway the demonstrated things confirmed that VC-1 is definitely not so adopted in higher end broadcasting (HD / SD DTV, IPTV, etc). Only mobile applications seemed to still get some interest for VC-1.

ps: I was wondering, are you the person from CMC ?

[Golgot13]

ps: I was wondering, are you the person from CMC ?

If I was a person of CMC, I will be pro-MS because I received MS support 6 or 9 month before all (alone in France...).


And I surprise to know that CMC was interested by H264 (they know you because you talked about them).
This company said that VC1 is the better codec... better than H264... but today they try to find H264 solution...

[diogen]

I lost count how many times local experts (non-MS) proclaimed AVC being so much better than VC-1 that there is no contest. The only thing that I don't get is this: why repeat it so often when only a stupid won't get it by now. I can't believe convincing those few MS employees (that post here) can be of such importance to these know-it-all experts.

Diogen.

[Sergey A. Sablin]

I lost count how many times local experts (non-MS) proclaimed AVC being so much better than VC-1 that there is no contest. The only thing that I don't get is this: why repeat it so often when only a stupid won't get it by now. I can't believe convincing those few MS employees (that post here) can be of such importance to these know-it-all experts.

Diogen.

+1

Sergey.

[Golgot13]

I lost count how many times local experts (non-MS) proclaimed AVC being so much better than VC-1 that there is no contest. The only thing that I don't get is this: why repeat it so often when only a stupid won't get it by now. I can't believe convincing those few MS employees (that post here) can be of such importance to these know-it-all experts.

Diogen.

+2
Industry understood this many year ago...


But MS with communication (and more...) show that VC1 is better:
http://www.microsoft.com/windows/windowsmedia/howto/articles/vc1techoverview.aspx#VC1ComparedtoOtherCodecs

[Sharktooth]

As every codecs VC-1 has its own pros and cons. At high bitrates it's not that bad, expecially when it comes to retaining fine details.
However all codecs are converging to transparency at high bitrates... but VC-1 seems to be particularly tuned to reach a very good quality at HD-DVD/BD typical bitrates.
Those bitrates are so high that the difference between VC-1 and h.264 can be no longer spotted.
So, for this kind of usage, VC-1 is a good codec, but affirming it's generally comparable or that beats h.264 is just a big fat lie.

[Golgot13]

VC-1 is a good codec, but affirming it's generally comparable or that beats h.264 is just a big fat lie.

Agree, it's a lie.


I don't understand why we don't have lot of HDDVD or BD in H264.
HDDVD use VC1 because VC1 encoder was free, but not free today... ???
BD from Sony use MPEG2 because they bought MPEG2 HD encoder or it' realtime encoder ?

[Inventive Software]

Toshiba has interests in H.264, which also helped develop HD-DVD. VC-1 is primarily base-lined off WMV9, which is MS, which supports HD-DVD. Sony use MPEG-2 because they have THAT much pride, and so little interest in the consumer. ;)

[Sagittaire]

Toshiba has interests in H.264, which also helped develop HD-DVD. VC-1 is primarily base-lined off WMV9, which is MS, which supports HD-DVD. Sony use MPEG-2 because they have THAT much pride, and so little interest in the consumer. ;)

Well MPEG2 in 25-35 Mbps interval is really not a bad choice.

[Inventive Software]

Compared to H.264 or VC-1 at 15-25 Mbps, perhaps so. But for easy encoding, MPEG-2 is the first way to go for most small-time movie companies that want to switch from DVD to HD-DVD, since it doesn't necessarily mean learning a new software suite.

One point that I've noticed with most H.264 / VC-1 / MPEG-2 codec comparisons in HD, is that the MPEG-2 version is *usually* a bit softer on the eye. But that's the only consistent thing IMHO.

[diogen]

Industry understood this many year ago...Really? How many years ago?
My understanding is AVC is just few years old...
And it was a superior codec the first year maybe on paper...

I can't understand why the argumentation is shifted at will depending on availability of good points.

Industry acceptance.
When comparing codec capabilities, who the f*ck cares about industry acceptance?
Does it actually need to be proven that politics plays at least as much a role in codec decisions than the codec capabilities?
What about workflow? Not being spoiled with access to commercial encoders, I find MS Expression A/B comparison feature
of higher value to me - actually showing what all the myriad parameters mean - than any features command line encoders can offer.

Codec performance.
ED was a good test with a non-representative source. Everybody agrees that at 25Mbps+ codec doesn't matter.
Real life non-CG material (everybody is expecting Ben to provide) should answer the question about grain preservation and such...
Why not wait until this is done before beating the drum of AVC's superiority?

Diogen.

[foxyshadis]

Compared to H.264 or VC-1 at 15-25 Mbps, perhaps so. But for easy encoding, MPEG-2 is the first way to go for most small-time movie companies that want to switch from DVD to HD-DVD, since it doesn't necessarily mean learning a new software suite.

One point that I've noticed with most H.264 / VC-1 / MPEG-2 codec comparisons in HD, is that the MPEG-2 version is *usually* a bit softer on the eye. But that's the only consistent thing IMHO.

You have to buy the HD package of whatever you use anyway, so as long as it's out there you might as well buy the H.264 or VC-1 package, unless it costs a great deal more. I can't imagine the interfaces of any of them being very different, aside from advanced codec options.

[Golgot13]

Really? How many years ago?
My understanding is AVC is just few years old...

I can give a simple example at professional people, NAB2005 and NAB2006:
NAB2005, lot of companies talked about VC1 and promised VC1 hardware (there was some H264 solution)
NAB2006, all companies showed H264 solutions and very few VC1 solutions

About H264 the first HD channel (not SD) broadcasted in 2005 and the first HD package (Voom) in 2006.


And it was a superior codec the first year maybe on paper...
You can read this link, VC1 is better but we saw nothing about measure and nothing about other encoder use on test:
http://www.microsoft.com/windows/windowsmedia/howto/articles/vc1techoverview.aspx#VC1ComparedtoOtherCodecs


VC1 is good codec on small CPU computer (like WMVHD) and at 20-25Mbps can make good picture quality
with movie grain (with very few preprocessing).

[zambelli]

You can read this link, VC1 is better but we saw nothing about measure and nothing about other encoder use on test:
http://www.microsoft.com/windows/windowsmedia/howto/articles/vc1techoverview.aspx#VC1ComparedtoOtherCodecs
I'm sure everybody read it one of the previous 6-7 times that you posted that link. Give it a rest. Yes, there are dubious references in there to "internal tests" of MPEG-2 that I agree should be removed, and though the H.264 references are more specific (DVD Forum's 2002 evaluation) I think they are too old to be in that article at this point. I'll work on having that cleaned up as I don't particularly like seeing marketing mix with technical articles either.

I saw last bad new for MS today in EU. I hope you will not have same problem with VC1 toolkit
given only at some specific companies 6 or 9 month before others.
:eek: Are you serious? Umm, you do realize there's nothing illegal about a company choosing who to do business with based on cost effectiveness, right? If every time a company entered a business partnership it got sued for NOT having partnerships with every other company in the business, every company would be bankrupt by now.

I lost count how many times local experts (non-MS) proclaimed AVC being so much better than VC-1 that there is no contest. The only thing that I don't get is this: why repeat it so often when only a stupid won't get it by now. I can't believe convincing those few MS employees (that post here) can be of such importance to these know-it-all experts.
Most of the time when I get involved in these VC-1 vs H.264 threads I'm primarily ensuring that comparisons of Microsoft VC-1 and other encoders are being done fairly and that VC-1 information is correct and unbiased. I don't have a problem with anybody putting together a fair test and making conclusions of their own. Why exactly certain people feel the need to prove H.264 superiority or demand proof from me about VC-1 superiority when I have in fact never stated that H.264 is an inferior or bad codec - is beyond me too.

[diogen]

I can give a simple example at professional people, NAB2005 and NAB2006:
NAB2005, lot of companies talked about VC1 and promised VC1 hardware (there was some H264 solution)
NAB2006, all companies showed H264 solutions and very few VC1 solutionsFRExt was finished by the end of 2004.
There couldn't have been much of H.264 solutions by April 2005 (NAB2005).
And according to Tom McMahon from Broadcom, without FRExt H.264 wasn't much better than MPEG-2.
You can read this link, VC1 is better but we saw nothing about measure and nothing about other encoder use on test:
Good point.
So lets not follow Microsoft in this respect and just claim AVC is better (gazillion times) but offer some proof (measure - in your words).

Diogen.

[Golgot13]

So lets not follow Microsoft in this respect and just claim AVC is better (gazillion times) but offer some proof (measure - in your words).

Yes, you're right but I posted in January on this forum that my company have VC1 toolkit
and I would do comparison. But all company, after, received email of MS which said it's prohibited
because they signed NDA. Today my company don't have a VC1 license (we work with other company for this)
we can have last version (we will test it).


I post a on other thread at Ben if I can use the last version (when we receive it) to make comparison
but he don't said yes or no (there is some NDA to use it).

[Golgot13]

:eek: Are you serious? Umm, you do realize there's nothing illegal about a company choosing who to do business with based on cost effectiveness, right? If every time a company entered a business partnership it got sued for NOT having partnerships with every other company in the business, every company would be bankrupt by now.


In EU, the "unfair competition" is prohibited

[obieobieobie]

Oh, come on, Golgot13. Choosing ones business partners is not unfair competition.

(I'm from Europe)

[Golgot13]

Oh, come on, Golgot13. Choosing ones business partners is not unfair competition.
(I'm from Europe)

No, when you give a software for free at one company (or 2 only) and help to make a product, you lock the market.
Because this software was alone to make sure a product compliant with specification (others software worked at the end of 2006).

And MS choosed only two partner only in Europe about this.
All other company which asked to have this software (without support because MS can't support much company)
received no answer (I'm not alone for this).

Some companies lost time and money (no contract because they can not produce HDDVD).
And HDDVD format lost too, because they was only two companies (with MS support)
which can produce HDDVD in Europe (Q2-Q3 2006). Today it's better because MS made Technical Training
to have more company which could make HDDVD (so in Europe, there are more HDDVD project).

[Golgot13]

FRExt was finished by the end of 2004.
There couldn't have been much of H.264 solutions by April 2005 (NAB2005).
And according to Tom McMahon from Broadcom, without FRExt H.264 wasn't much better than MPEG-2.

You're right, but FRExt is for professional application (like Studio Profile with MPEG4 Part2).
Today, H264 used by TV Broadcaster are in 8bit and 4:2:0 (Last year in Main Profil and this year in High Profil).
And it's same for HD disc and box (IPTV).

Do you know if there is some chipset with support "High 10", "High 4:2:2" and "High 4:4:4" ?

I sure when it will be available we will see some new professional HD VTR
(Sony and Panasonic use today FPGA on Camcorder for Intra at High 10).

[diogen]

I have to say upfront, I'm not privy to any inside information.
This compression thing is my hobby and has less than nothing to do with my day job.
Do you know if there is some chipset with support "High 10", "High 4:2:2" and "High 4:4:4" ?No, I don't.
But reading the insiders' posts on AVS I got the impression there is not much to be gained abobe 4:2:2/10bit.

You're right, but FRExt is for professional application (like Studio Profile with MPEG4 Part2).My understanding is FRExt Extensions make AVC the great codec it is today and are applicable to 4:2:0/8bit (what BD/HD use) as well.

BTW, does x264 use some features that are from FRExt domain?

Diogen.

[diogen]

I have a VC1 encoder to make my test but I can not publish C'mon, Golgot13.
If you have some internet/forum posting experience, you should know how to make public the results of your encodings without risking them being traced back to you.

Diogen.

[akupenguin]

You're right, but FRExt is for professional application (like Studio Profile with MPEG4 Part2).
FRExt is a collection of tools. Some of them are like studio profile (extra colorspaces and bit depths), others are useful for all applications (adaptive dct size and cqm).
8x8dct really should have been in baseline profile.