How Will High Bit Rate VC-1 Compare to AVC

[Swede]

Originally Posted by reiella

By the same token, just because the bit-rate is higher, doesn't mean the image is benefiting from the increased bandwidth.

The metrics facilitate the result, and the result should be what we care about, not the factors going into it.

Actually, Yes it does.

Let me put this as simple as possible. Everything else being equal: source, master, encoding tools and settings, a lower bitrate video will NEVER look better than a higher bitrate one, AND the higher bitrate video will ALWAYS have MORE DETAIL than the lower bitrate video.

How can I say this with certainty? It's very simple really, because if the bitrate had already been maxed out with the lower bitrate, a higher bitrate would not be possible, since the codec would not be able to use it. Now, in SOME cases the higher detail available with a higher bitrate may not be that noticable, but it IS there. Likewise, in many cases the differences can be very noticable, especially on hard to encode, bitrate starved scenes.

It's the same principle as with jpg images, you may or may not notice the difference between different quality settings, but if the file size is lower, the detail level will be lower too.

[PaulGo]

For those that are interested, I got an older thread restarted on AVS:

"Codec Wars" : The attempt of an objective AVC/h.264 versus VC-1 benchmark

http://www.avsforum.com/avs-vb/showthread.php?t=883667

Also, some real technical experts from Doom9 are participating this may get very interesting!

[reiella]

Originally Posted by Swede

Actually, Yes it does.

Let me put this as simple as possible. Everything else being equal: source, master, encoding tools and settings, a lower bitrate video will NEVER look better than a higher bitrate one, AND the higher bitrate video will ALWAYS have MORE DETAIL than the lower bitrate video.

How can I say this with certainty? It's very simple really, because if the bitrate had already been maxed out with the lower bitrate, a higher bitrate would not be possible, since the codec would not be able to use it. Now, in SOME cases the higher detail available with a higher bitrate may not be that noticable, but it IS there. Likewise, in many cases the differences can be very noticable, especially on hard to encode, bitrate starved scenes.

It's the same principle as with jpg images, you may or may not notice the difference between different quality settings, but if the file size is lower, the detail level will be lower too.

Which of these three images has the higher picture quality?







The point, I was aiming to make is that there are ways to have No Gain from increased bandwidth. Not negative gain, but No Gain.

[ Add / Edit ]
And I'm sorry for derailing it like this PaulGo, pretty interested in the codec tests, just frustrated at the side comments.

[UTVOL06]

Originally Posted by PaulGo

I guess it goes with Amir's "good enough" philosophy.

I have read into the AVC vs. VC-1 debate and what I found was that at low bitrates AVC held up much better and offered more detail.

At high bitrates AVC/VC-1 looked virtually identical and arguments which was better at this level of bitrate was trivial or academic.

So in general I would prefer AVC, but if studios offered only high bitrate encodes VC-1 would be just as good.

Just look at Shoot'em Up. Its a medium to highbitrate encode thats too high for HD players to support. The sharpness and level of facial detail is incredible in it.

[Branden]

Originally Posted by UTVOL06

I have read into the AVC vs. VC-1 debate and what I found was that at low bitrates AVC held up much better and offered more detail.

if true then that's interesting, because given HD-DVD's lower bandwidth capabilities and space restrictions you'd think AVC would be much more common on it.

Originally Posted by UTVOL06

At high bitrates AVC/VC-1 looked virtually identical and arguments which was better at this level of bitrate was trivial or academic.

agreed. we should be more concerned about whether studios are willing to utilize available bandwidth on blu-ray rather than which codec is used.

[Teazle]

Originally Posted by Neo65

When the source is lossy compressed video like mpeg2, there is already significant loss of detail from the original, so you can't create more actual entropy than what is available in your mpeg2 source, you can at most create more noise/artifacts but by transcoding with a different codec, hence different DCT, different math, you're getting even further away from the original master.

(I hope this is not getting too OT.) Re: the use of a lossy source like mp2. When I started ripping DVDs to mp4s I was worried that PQ would be awful for that very reason. In particular I expected that the picture couldn't compare to professionally-produced mp4s on UMD, since there the compressionist presumably has access to the uncompressed master or at least to a much less heavily compressed version than a DVD-grade 720x480 mp2 with a max bitrate of 10mbps and an average bitrate much lower than that.

Yet in practice I find (or think I find) that my homebrewed mp4 rips sometimes come out looking better than the UMD mp4s. Playing back from flash memory, I have the advantage of being able to exceed 1.8 GB per movie (the limit of a UMD disc) and so can use a high fixed bitrate of up to 3Mbps, a higher CBR than the pro rips. (Like the UMDs I also rip to 720x480 mp4 AVC using main profile 3.)
** Edit-- the professionally produced versions almost certainly don't use fixed bitrate!

I have viewed about 40 UMDs and have home rips that look better than any of them. When I get round to it I'm going to do a direct comparison using the DVD version of one of the titles I own on UMD.

Is it counterintuitive that a high bitrate video from a lossy low-res source should look better than a slighly lower bitrate video from a lossless higher-res source? The simple program I'm using (psp video 9) doesn't let me look under the hood and see exactly what's going on behind my profiles, and even if it did, as a beginner I probably wouldn't understand it.

[patrick99]

Originally Posted by Branden

agreed. we should be more concerned about whether studios are willing to utilize available bandwidth on blu-ray rather than which codec is used.

Absolutely, that is the key question, together with whether studios want to put in gimmicky "bonus" features when the consequence is to compromise PQ and AQ on the movie itself.

[reiella]

Originally Posted by Teazle

(I hope this is not getting too OT.) Re: the use of a lossy source like mp2. When I started ripping DVDs to mp4s I was worried that PQ would be awful for that very reason. In particular I expected that the picture couldn't compare to professionally-produced mp4s on UMD, since there the compressionist presumably has access to the uncompressed master or at least to a much less heavily compressed version than a DVD-grade 720x480 mp2 with a max bitrate of 10mbps and an average bitrate much lower than that.

Yet in practice I find (or think I find) that my homebrewed mp4 rips sometimes come out looking better than the UMD mp4s. Playing back from flash memory, I have the advantage of being able to exceed 1.8 GB per movie (the limit of a UMD disc) and so can use a high fixed bitrate of up to 3Mbps, a higher CBR than the pro rips. (Like the UMDs I also rip to 720x480 mp4 AVC using main profile 3.)

I have viewed about 40 UMDs and have home rips that look better than any of them. When I get round to it I'm going to do a direct comparison using the DVD version of one of the titles I own on UMD.

Is it counterintuitive that a high bitrate video from a lossy low-res source should look better than a slighly lower bitrate video from a lossless higher-res source? The simple program I'm using (psp video 9) doesn't let me look under the hood and see exactly what's going on behind my profiles, and even if it did, as a beginner I probably wouldn't understand it.

It can vary. My experience was the reverse with ATHF, although that was admittantly with a mp2 [and a 1gb memory stick, although, I was comparing a single 15minute episode]. Was getting some really bad desynchs with audio [would fix if I paused etc]. Prolly alot better now though, especially as I imagine PSPVideo 9 is a bit better now than back near launch . I might have to give it another shot.

[PaulGo]

This is from AVS I was responding to a question and Richard Casey had a interesting response.

Originally Posted by PaulGo;

To me (and I am not by any means an expert) it would seem the higher the bit rate the less work a compressionist would have to do since in theory we would be getting closer to the actual product. The test I would want to see is how well the codecs could do without any special tweaking and then perhaps a set amount of tweaking - to be determined by the experts as to what is prudent. You do bring up some very good points which need to be worked out, since it is how closely the final product matches the original.

Originally Posted by RBFilms;

That is correct.

BD has 48mbps compared to HD which only has 30mbps of bandwidth to work with.

The more bandwidth the less one has to molly coddle the encode. It is actually more efficient to use higher bandwidth rates. The higher bit rates are also an advantage dependent upon which CODEC you are using.

AVC tends to perform better given more bandwidth and is the preferred choice of many authoring houses that we work with given the opportunity to use higher bit rates.

VC-1 is probably the better choice for limited bandwidth applications. This may be due to the fact that is was designed for maximum performance in lower bit rate environments.

Again, this is not just my opinion ... but the opinion of several highly respected authoring labs, compressionists, and engineers we work with.

Everyone has an opinion, especially in the AVS Forum. However, I have spent 34 years in the industry working with the very best ... the elite so to speak ... and while we do not agree on everything ... we all agree on most things.

Having produced over 1,000 intellectual properties and worked on the development, distribution, sales, and marketing of over 5,000 programs during my career. I can honestly say I have seen, heard, witnessed, and dealt with enough to make a valuable contribution that is as good as any other opinion on this board.

The big difference between others and myself is that I don't assume I know any professional's experience level, work history, or background nor do I speak disrespectfully about them or their work experience. This is especially true if I have not seen their resume or know their entire body of work and book of business.

It seems some people here like to go off without have a clue about the facts and spew misinformation in an attempt to spread wholesale propaganda .... for reasons they believe will help them gain in the end.

All I can say is this. I am more motivated now than ever before to assure people know the truth and have what they need to decide for themselves what is best.

In the end, the truth almost always wins out ... just like it has in the past and is playing out now with technology that is being accepted by the majority of industry professionals and consumers alike.

Again, both VC-1 and AVC Codecs may serve a purpose in the end. Both do a good job. Neither of them are bad. However, I personally see advantages in using AVC for specific content and for certain when I have over 20mbps to work with.

However, that is just my opinion and the opinion of at several dozen other industry professionals. You do not have to believe any of us. Just look at our Encoder "Crash Test" Shootout on one of our upcoming releases and decide for yourself. At the very least, I think we will all agree on one thing in the end....that there are differences between the codecs.

[Cyorg]

Originally Posted by PaulGo

because of Amir's ranting it got locked. .

I was under the impression that he doesn't work for M$ anymore. Why does he still say "We" when he's talking about M$?
I'm probably wrong, so forgive me if I am.

[reiella]

Originally Posted by Cyorg

I was under the impression that he doesn't work for M$ anymore. Why does he still say "We" when he's talking about M$?
I'm probably wrong, so forgive me if I am.

He's referencing the past tense I believe. He was part of the 'we' at the time when alot of the development happened on the vc-1.

[Grubert]

Kris Deering just posted this on AVSforum:

I don't think we'll be seeing long term support of VC-1 going forward. MS delivered the last tools for VC-1 already and has moved away from the codec in terms of future support from what I've heard.

This could be big news.

http://www.avsforum.com/avs-vb/showthread.php?t=996498

[PaulGo]

Interesting discussion on AVS:

From:Richard J. Casey R&B Films, Ltd.

Thanks for the feedback. I found that info online.

http://dostudio.netblender.com/wikipapers/Encoders1.asp

Nature's Journey was an early HD-DVD / BD release and we used the latest and greatest and most current version of VC-1 that was available at the time. Technicolor did the encodes.

Honestly, I understand your need to promote CODEC and Technology as part of your obligation to your employer ... and your commitment to the development team.

I have no such affiliations, obligations, or loyalty to be concerned with. We honestly see diffrences between them ... as I am sure you yourself would admit ... otherwise ... one would not be any different from the other. For anyone to claim VC-1 is better immediatley confirms there is a difference and leaves the discussion open to debate regarding AVC verses VC-1.

I am not here to debate as I am already sold on AVC and so are all of the authoring houses I have spoken with.

VC-1 is good ... and I would use it in some applications depending on content.

I find AVC ... compared to the flavors of VC-1 I have tried .... looks sharper and more detailed, handles blacks and shadow detail better, and appears to take better advantage of the additional bandwidth I tend to give my encodes.

These are only my observations. I am not here to promote or bash one or the other. However, I do have some special edition High Bit Rate releases coming out in 4th quarter and I will be using Tiger.

I woudl be happy to do test encodes using the latest version of VC-1 and do a new comparison to Tiger. If you can help me cooirdinate something like this I will be happy to revisit AVC verses VC-1.

I would even be willing to put the test encodes on any of my upcoming releases as an Easter Egg for AVS Forum fans.

Who can you recommend for a test encode at 36mbps to 38mbps using VC-1 with our content? WHo is the best encoding technician you know? I will be happy to give this a try.

Quote:
Originally Posted by benwaggoner
The SDK is an implementation.


Scenarist is an authoring tool. Do you mean CineVision? Original flavor or PSE? What version? You'll get massive differences in VC-1 output between Cinevision and Cinevision PSE. And there are a very wide array of improvements between the original PSE release and the current one being used for many Blu-ray titles today.

As far as I know, your original disc is the only widely released title done with the non-PSE Cinevision VC-1 encoder, so is not representative of any other VC-1 HD DVD or Blu-ray titles.


This is something that the Hollywood compression houses have always felt was something that our tools did extremely well. So what are they comparing Tiger to?


Ah. If that's the only one you tried, than you haven't been exposed to the VC-1 that is being used for actual discs.

Right now, we like Tiger. Even the $300,000 "dumbed down" version of Tiger ... the Tiger Cub .... incorporates some of the most important technology only available with Tiger ... so I have been told.

Rumor also has it that Sony's new AVC Encoder uses some proprietary Tiger technology. Tiger is not for sale but would probably cost a fortune if it was.

We are using the the proprietary Tiger Encoder at Technicolor for all of our future projects at this point. It appears to do a great job.



Where did you find THAT? It's not helpful to consumers, it's just factually wrong.

"There are numerous licensing entities involved in the VC-1 standard which has lead to the greatest number of encoders falsely claiming BD output"?
The sole licensing entity for VC-1 is MPEG-LA, the same entity that does MPEG-2 and H.264.

And encoders falsely claiming BD output? I've never even heard a rumor about that.

"...very few Blu-ray titles have used this codec"?
Manifestly false. All three codecs have seen quite a bit of use, and there are certainly many more new titles using VC-1 than MPEG-2.
__________________
Richard J. Casey
---------------------
R&B Films, Ltd.
Producers Guild of America, New Media Council

http://www.avsforum.com/avs-vb/showt...3#post14114923

[Entertainment72]

Was Shoot Em Up VC-1? I thought it had great PQ.... a 4.8/5!

[WickyWoo]

Yes it was.

There's nothing wrong with VC-1. But as systems are replaced, fully expect to see its use wane since MS is no longer actively developing it

[Neo65]

All else being the same, the CABAC entropy encode (lossless compression in the last stage) of AVC still beats VC-1's use of VLC. Remember that CABAC is a form of lossless compression based on arithmetic coding. VLC is based on huffman coding. It take some familirity to understand that unlike lossy compression comparisons (which is so subjective that people can play tricks till the cows come home to make outlandish claims that it's a waste of time to play that game) one form of lossless compression can be proven to be superior the same way (for example, you take a file, compress using zip and rar and look at the resulting file size --- across a wide sample of inputs, you can then draw your own conclusions).

In a large sample of streams, CABAC yields a 10-15% size savings over CAVLC.

As for the internals of the codec itself, in HD resolution, VC-1 advanced profile is better than AVC's Main profile, but h.264 has a HighProfile FRExt. And h.264 HighProfile (with 8x8 transforms) imo is just a little bit better than VC-1 if we ignore the entropy encode side and stop at syntax elements.

The world has moved to h.264 in broadcast TV (check what all european, asian, south american and north america standards bodies controlling TV broadcast have mandated as the replacement for MPEG2) and non-ms supported iptv. Other than being less processor intensive, there is no reason for VC-1 or WMV9 to live. Even microsoft no longer has a team that actively supports further work on it.

The point really is that it is difficult to get onto the technical argument and it really is a waste of time since that bar can move so easily. You bring up PSNR or JND error, and people can say subjective tests are better. You try to organize a subjective test and people try to game the process by playing with selection of inputs that perform better on one codec, you try to widen the number of streams and people ask for more time to wait for the other stream before they optimize their own encode to beat codecA. Then codecA can be further optimized to beat codecB's implementation.

In the end, it's close enough technically that it won't matter. FWIW, VC-1 syntax elements are simpler and smaller than AVC's even if their entropy encoding is weaker. But noone cares about this stuff.

In the end, the reality is that there is no future for VC-1 --- the template for this is .ARJ files, it never really kept pace with .ZIP, and it died out, but now .RAR is giving .ZIP some competition.

Other than companies forced (or paid) to build a decoder for it to handle some esoteric applications that has VC-1, noone if going to do any R&D on building a chip to do VC-1 compression. We will never see a HD camcorder that uses VC-1, we will not see broadcast equipment that uses hardware VC-1 compression --- no chip exists and no chip ever will support VC-1 compression in hardware.

[thejoeman2]

So, I take it we will never see VC-2. Microsoft left the race and Google took over with VP9.

[Wendell R. Breland]

Originally Posted by thejoeman2

So, I take it we will never see VC-2. Microsoft left the race and Google took over with VP9.

Actually the dominating video codecs is AVC (H.264) and HEVC (H.265). Blu-ray and UHD Blu-ray use these, no VP9. VP9 has a replacement in AV-1, VVC (H.266) will most likely be used if 8K should ever gain any traction.