Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

 

Go Back   Doom9's Forum > Video Encoding > MPEG-4 AVC / H.264

Closed Thread
 
Thread Tools Search this Thread Display Modes
Old 31st October 2008, 16:21   #41  |  Link
LoRd_MuldeR
Software Developer
 
LoRd_MuldeR's Avatar
 
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 12,921
I'm pretty sure we'll see new video compression algorithms, either further optimiztions of existing approaches or completely new approaches or a combination of both. But still the question: How long will it take until we see something that is really far superior to H.264? And now that H.264 has just been widely adopted in the industry and by customers (including "mobile" and Internet), it may take a long time to establish a completely new format. The SD to Full HD step was a huge one and it made H.264 necessary, because MPEG-2 wasn't really handy for Full HD footage. But are we going to see another big step, that makes the "next generation" video compression necessary, anytime soon? I'm not so sure about that, but we'll see, of course...
__________________
There was of course no way of knowing whether you were being watched at any given moment.
How often, or on what system, the Thought Police plugged in on any individual wire was guesswork.



Last edited by LoRd_MuldeR; 31st October 2008 at 16:26.
LoRd_MuldeR is offline  
Old 31st October 2008, 19:28   #42  |  Link
Soulhunter
Bored...
 
Soulhunter's Avatar
 
Join Date: Apr 2003
Location: Unknown
Posts: 2,812
My prediction for the next 5 years... I'm with LoRd_MuldeR for the thing with video compression -> Progress in video compression algorithms will be kinda slow, and x264 (or at least some other h.264 based encoder) will still be the winner of Doom9's codec comparisons... But regarding "higher resolutions than Full HD" -> UHDTV is already just "around the corner" in Japan! Probably there will be some test broadcasts in 5 years, and UHD capable (tho very very expensive) displays... Another 5 years and a lot of us will say "my new UHD movies look sooo much better than the old HD ones". HD might be "wowing" today, but remember, there was also a time where VHS or 16bit VGA res. was impressive! ;]
__________________

Visit my IRC channel

Last edited by Soulhunter; 31st October 2008 at 22:14.
Soulhunter is offline  
Old 31st October 2008, 20:25   #43  |  Link
prOnorama
Registered User
 
Join Date: Mar 2005
Posts: 249
Quote:
Originally Posted by mikeytown2 View Post
Prediction: Most people will not be buying their music in a physical format by then.
Outcome: iTunes has taken over for the most part.
That's nonsense CD's are still sold very much. It's true that the CD single was basically killed by online music sales (like iTunes and others), but that's for single tracks only.
prOnorama is offline  
Old 31st October 2008, 21:22   #44  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,405
My thoughts are that MPEG-4 will be around for a very long time. UHD will take awhile to get any sort of traction (I'm thinking 10-15 years). HDTV has been around in one form or another since the '80s, and is just within the last 3-4 years seeing widespread adoption. Heck, the Russians had a 1125 line analog system as early as 1958 for military applications!

http://en.wikipedia.org/wiki/Hdtv

Anyway, compression technology is constantly improving (look at x264 over the last year!!), though I have a hard time imagining what the next "big thing" will be. Available bandwidth and storage capacity continue to increase at astonishing rates (well... storage at least), which lessens the need for super awesome compression.

I imagine by the time we need to transmit / store UHD on any sort of scale, something new will be developed to fit these needs. I wonder if something intermediate, a-la 4K (RED camera, anyone?) will fill the void.

I'd like to see a trend towards quality - i.e. higher bit depth / wider color gamut. This mostly means better displays. Unfortunately, most people either don't care, or can't tell a difference (realistically the same thing).

Who knows... I'm pretty damned happy with BluRay for now... especially my half-size backup MKVs!

~MiSfit

Last edited by Blue_MiSfit; 31st October 2008 at 21:32.
Blue_MiSfit is offline  
Old 31st October 2008, 23:05   #45  |  Link
Sagekilla
x264aholic
 
Join Date: Jul 2007
Location: New York
Posts: 1,752
I agree with Blue_MiSfit here: I'd like to see the trend towards higher bit depths in displays and for the actual content. Maybe soon we'll move towards YV12 with full chroma resolution? I remember reading an article once that said compressing an image at 4:2:0 sampling vs 4:4:4 sampling didn't make a huge difference in final file size, but the 4:4:4 file could give higher potential quality because of the larger resolution to start with, as well as some better prediction available from having all samples.

But, that was an article I read from God knows where. Is it true? I personally have no idea, but I'd be interested in seeing if it does in any lossy compression scheme.
__________________
You can't call your encoding speed slow until you start measuring in seconds per frame.
Sagekilla is offline  
Old 31st October 2008, 23:54   #46  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,405
YV12 can never have full chroma resolution It's 4:2:0. 10 bit would be a much better improvement IMO - that way we wouldn't have to hide banding with dither/noise.

~MiSfit
Blue_MiSfit is offline  
Old 31st October 2008, 23:59   #47  |  Link
LoRd_MuldeR
Software Developer
 
LoRd_MuldeR's Avatar
 
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 12,921
Quote:
Originally Posted by Blue_MiSfit View Post
YV12 can never have full chroma resolution It's 4:2:0. 10 bit would be a much better improvement IMO - that way we wouldn't have to hide banding with dither/noise.

~MiSfit
Okay, this may sound stupid, but what exactly do the numbers in x:y:z stand for and how does YV12/YUY2 fit in?
__________________
There was of course no way of knowing whether you were being watched at any given moment.
How often, or on what system, the Thought Police plugged in on any individual wire was guesswork.



Last edited by LoRd_MuldeR; 1st November 2008 at 00:07.
LoRd_MuldeR is offline  
Old 1st November 2008, 00:18   #48  |  Link
STaRGaZeR
4:2:0 hater
 
Join Date: Apr 2008
Posts: 1,307
Quote:
Originally Posted by LoRd_MuldeR View Post
Okay, this may sound stupid, but what exactly do the numbers in x:y:z stand for and how does YV12/YUY2 fit in?
http://en.wikipedia.org/wiki/Chroma_subsampling

Hope that helps
__________________
Specs, GTX970 - PLS 1440p@96Hz
Quote:
Originally Posted by Manao View Post
That way, you have xxxx[p|i]yyy, where xxxx is the vertical resolution, yyy is the temporal resolution, and 'i' says the image has been irremediably destroyed.
STaRGaZeR is offline  
Old 1st November 2008, 02:50   #49  |  Link
akupenguin
x264 developer
 
akupenguin's Avatar
 
Join Date: Sep 2004
Posts: 2,393
Quote:
Originally Posted by zeeman_88 View Post
In my humble opinion, the current motion estimation algorithm of using trial and error till findings the “best” motion vectors is not that efficient!
There are encoder-side alternatives to trial-and error, such as Skal's analytic subpel search, or Tuukka's transform-based exhaustive search. But those are a question of implementation algorithm, not a change in the compression standard. And I have yet to find one that's unambiguously better or faster than trial-and-error with SIMD optimizations.

Quote:
One way to do that is the 3D de-correlation transform… And may be there will be better way to do that.
Orthogonal 3D wavelet is much worse than mpeg1, it's approximately equivalent to hierarchical B-frames with every mv set to 0,0.
I know of exactly one attempt to capture some of the motion redundancy without actual motion vectors. This algorithm is much better than orthogonal 3D wavelet, and still much worse than mpeg1. As a bonus it's hella slow.
Motion compensated 3D wavelet works (approximately equivalent to normal hierarchical B-frames), but then you're hardly revolutionizing motion prediction.

Quote:
Right now the “image” and “audio” compression algorithms are more sophisticated than video algorithm. Basically video compression algorithms are still the same since the days of mpeg1.
That's why h264 intra-frames provide better quality-per-bitrate than any standard lossy still-image format, right?

Last edited by akupenguin; 1st November 2008 at 07:44.
akupenguin is offline  
Old 1st November 2008, 04:22   #50  |  Link
Caroliano
Registered User
 
Join Date: Feb 2005
Location: São Paulo, Brazil
Posts: 392
Quote:
Prediction: Various security measures will be integrated to provide anonymous downloading.
Outcome: Not really unless your a nut and using Tor for torrents.
You are forgoting japanese P2P programs. Winny, Share and Perfect Dark, in chronological order.

Quote:
Also I'm not sure if we will ever need much higher resolutions than "Full HD", simply because you can't integrate even bigger screens into your living room. Not everybody can have a separate room for TV and Beamer.
People (japaneses..) are already discussing SHV, with 8000 by 4000 pixels, and standart designers are thinking how to make better codecs for that: http://diracvideo.org/node/13

Not in 5, but maybe in 10 years we can start to paint our walls with hi-resolution OLEDs or E-INK instead of having small 50" LCD tvs like now.
Caroliano is offline  
Old 1st November 2008, 04:33   #51  |  Link
LoRd_MuldeR
Software Developer
 
LoRd_MuldeR's Avatar
 
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 12,921
Quote:
Not in 5, but maybe in 10 years we can start to paint our walls with hi-resolution OLEDs or E-INK instead of having small 50" LCD tvs like now.
However the field of vision is limited by the human eye. So there is some natural limit for screen size, unless you have a cinema hall at home
__________________
There was of course no way of knowing whether you were being watched at any given moment.
How often, or on what system, the Thought Police plugged in on any individual wire was guesswork.


LoRd_MuldeR is offline  
Old 1st November 2008, 04:53   #52  |  Link
Caroliano
Registered User
 
Join Date: Feb 2005
Location: São Paulo, Brazil
Posts: 392
I noted that animes are changing since 2000 for ever smaller characters in the screen, as you can see things more clearly in bigger HDTV screens. Maybe movies will adapt to use most of those monstrous resolutions and sizes. 1080p is far from being more resolution than our eyes can distinguish.
Caroliano is offline  
Old 1st November 2008, 05:11   #53  |  Link
zeeman_88
Registered User
 
Join Date: Oct 2008
Posts: 18
Akupenguin

Thanks for the through answer. I admire your knowledge regarding temporal redundancy research.

I know there are several attempts to use 3D wavelet for exploiting temporal redundancy and I know they haven’t produce better results nor they are practical.

However, I am anticipating one day “in the future” we will see a video compression algorithm that doesn’t use trial and error or macro-blocks coding. Instead it will analyze, process and encode the whole GOP as one data chunk. Even if there is such algorithm today, it would not be practical, at least at the decoder side. However as more cores will be added to the CPU, such algorithms will be more feasible. Until then h.264 and its extensions will still be the video compression champion.

Now regarding h264 intra compression as the best lossy still-image format. I am not aware of that and have not compared it to JPEG2000 myself. Have you run such a test? I would be interested too see these results.

My expectation for such a test is that they would give comparable quality for low to medium compression. But at high compression ratio (80 and above), I doubt h264 Intra will be able to keep up.

zeeman_88 is offline  
Old 1st November 2008, 05:42   #54  |  Link
Dark Shikari
x264 developer
 
Dark Shikari's Avatar
 
Join Date: Sep 2005
Posts: 8,690
Quote:
Originally Posted by zeeman_88 View Post
My expectation for such a test is that they would give comparable quality for low to medium compression. But at high compression ratio (80 and above), I doubt h264 Intra will be able to keep up.
No, it trashes JPEG-2000 at pretty much any bitrate. This isn't surprising, since JPEG-2000 is a terribly designed format that has singlehandedly tarnished peoples' expectations of wavelet formats for an entire decade.

Its also not surprising since spatial intra prediction is impossible in an overlapped-block format.
Dark Shikari is offline  
Old 1st November 2008, 05:58   #55  |  Link
Sagekilla
x264aholic
 
Join Date: Jul 2007
Location: New York
Posts: 1,752
Quote:
Originally Posted by Blue_MiSfit View Post
YV12 can never have full chroma resolution It's 4:2:0. 10 bit would be a much better improvement IMO - that way we wouldn't have to hide banding with dither/noise.

~MiSfit

Full 4:4:4 would be handy though


@Zeeman: We could make x264 encode on a per-GOP basis, with it outputting ridiculously large amalgam of all the frames from the start to end of the GOP, which would constitute one huge data block
__________________
You can't call your encoding speed slow until you start measuring in seconds per frame.

Last edited by Sagekilla; 1st November 2008 at 06:02.
Sagekilla is offline  
Old 1st November 2008, 07:25   #56  |  Link
akupenguin
x264 developer
 
akupenguin's Avatar
 
Join Date: Sep 2004
Posts: 2,393
Quote:
Originally Posted by zeeman_88 View Post
Now regarding h264 intra compression as the best lossy still-image format. I am not aware of that and have not compared it to JPEG2000 myself. Have you run such a test? I would be interested too see these results.
My expectation for such a test is that they would give comparable quality for low to medium compression. But at high compression ratio (80 and above), I doubt h264 Intra will be able to keep up.
At high resolutions but very low bitrates, jp2k fares better than h264 since jp2k can reasonably efficiently discard whole subbands (which is almost as good as downscaling, though it still wastes a few bits), while h264 block size maxes out at 16x16. Nevertheless, in such extreme cases both jp2k and h264 benefit from downscaling the image and compressing the lower resolution version to the same filesize. And once you do that, h264 wins again. (Granted I had to tweak the resolution manually, as x264 doesn't optimize it for you.)

IOW, yes h264 fails at sufficiently low bpp, but low bpp is always avoidable.

Last edited by akupenguin; 1st November 2008 at 07:29.
akupenguin is offline  
Old 1st November 2008, 07:46   #57  |  Link
Dark Shikari
x264 developer
 
Dark Shikari's Avatar
 
Join Date: Sep 2005
Posts: 8,690
Quote:
Originally Posted by akupenguin View Post
At high resolutions but very low bitrates, jp2k fares better than h264 since jp2k can reasonably efficiently discard whole subbands (which is almost as good as downscaling, though it still wastes a few bits), while h264 block size maxes out at 16x16. Nevertheless, in such extreme cases both jp2k and h264 benefit from downscaling the image and compressing the lower resolution version to the same filesize. And once you do that, h264 wins again. (Granted I had to tweak the resolution manually, as x264 doesn't optimize it for you.)

IOW, yes h264 fails at sufficiently low bpp, but low bpp is always avoidable.
And this is a general problem with the concept of wavelets.

The primary benefit of wavelets is more efficient coding of large-scale detail.

But, in most cases, almost all your bits are spent on small-scale, not large-scale, detail.

But what if you're encoding at a bitrate low enough that you don't store small-scale detail?

... then you might as well downscale anyways, and you lose the wavelet advantage again.
Dark Shikari is offline  
Old 1st November 2008, 07:56   #58  |  Link
Blue_MiSfit
Derek Prestegard IRL
 
Blue_MiSfit's Avatar
 
Join Date: Nov 2003
Location: Los Angeles
Posts: 5,405
What about a hybrid mode? Wavelets for large-scale detail, and blocks for low-scale detail?

Or have I had too much damned kool-aid?

~Misfit
Blue_MiSfit is offline  
Old 1st November 2008, 07:58   #59  |  Link
*.mp4 guy
Registered User
 
*.mp4 guy's Avatar
 
Join Date: Feb 2004
Posts: 1,350
Imo it should be noted that bpp sensitivity is directly related to transform size, big transform = more possible data removal, small transform = limited data removal, and large transforms don't work well for a variety of reasons, this includes wavelets, curvelets, magiclets or any other buzzwordlet transform. Or, put another way, extremely low bpp (extremely low being defined as a case where the compression standard being used does not in anyway utilize a large percentage of the available resolution) is retarded, you should always use a realistic resolution for your bitrate budget, a qcif blurry mess is just as good as a wuxga blury mess, assuming very low and equivelent datarates (better if you account for superior performance of standalone interpolation to wavelet subband interpolation.

@ds
1 what is so horrible about jpeg2k
2 give me an example of a significantly better wavelet image coder from the same time frame.

I'm not saying j2k isn't pretty useless, but I happen to think this is largely because wavelets are over rated.



.
.
.
.
(hooray parenthesis)
*.mp4 guy is offline  
Old 1st November 2008, 08:04   #60  |  Link
Dark Shikari
x264 developer
 
Dark Shikari's Avatar
 
Join Date: Sep 2005
Posts: 8,690
Quote:
Originally Posted by Blue_MiSfit View Post
@ds
1 what is so horrible about jpeg2k
2 give me an example of a significantly better wavelet image coder from the same time frame.

I'm not saying j2k isn't pretty useless, but I happen to think this is largely because wavelets are over rated.
I'm not very well-versed in JPEG2K, but from those who know more than me about it, I've been told almost universally that it was a badly designed standard--not just that wavelets are problematic, but that the standard was just poor to begin with.

You'd have to talk to someone who's more familiar with the specifics to know exactly why.
Dark Shikari is offline  
Closed Thread

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +1. The time now is 00:21.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2018, vBulletin Solutions Inc.