backtop


Print 26 comment(s) - last by someguy123.. on Jan 6 at 6:17 PM

VP9 codec will also help reduce the data needs for HD video streaming

YouTube has announced that it will unveil a new 4K video streaming service at CES 2014. The 4K streaming content will use a new royalty-free codec developed by Google, which is called VP9.
 
Google has a list of 19 hardware partners that will support its VP9 codec take off for 4K streaming including some chipset vendors like ARM, Intel, Broadcom, and Marvell.
 
Sharp, Samsung, and Toshiba are also among the supporters. “This certainly isn’t a war of the video codecs," said Francisco Varela, global platform partnership director at YouTube.
 
If you don’t plan on adopting a 4K TV to enjoy the higher resolution streaming any time soon, VP9 still has some benefit for you. According to Varela, the VP9 codec will allow YouTube to reduce the amount of data needed to stream normal HD content by half. That means you can stream YouTube videos without as large an impact on your monthly data allotment.
 
VP9 hardware decoding will hit PCs and mobile devices first according to Varela, with TVs supporting the format hitting the market by 2015.

Source: Gigaom



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

The state of VP9
By bug77 on 1/3/2014 11:05:34 AM , Rating: 3
Are there publicly available encoders yet? Last I checked, many features (profile 1) were still to be implemented.
Also, if anyone can link a more recent comparison to h264, I'd really appreciate it. Past comparisons I've seen were rather narrow in scope and placed VP9 either in front or behind h264.




RE: The state of VP9
By robertgu on 1/3/2014 1:24:27 PM , Rating: 4
I don’t have a recent study directly comparing H.264 to VP9 but I found one comparing HEVC (H.265), the next-gen standard codex to the last-gen x264 and to Google’s next-gen VP9 codex.

According to the study completed in Dec. 2013, HEVC (H.265) is considerably more data efficient for same objective PSNR video quality in comparison to VP9. VP9 needed to increase bit rates by 79.4% to match HEVC's (H.265) objective video quality and x264 needed a bit rate increase of 66.4% to match HEVC’s objective quality. You can infer that since VP9 needs a larger bit rate increase to match HEVC’s quality vs. x264 needing less (79.4% increase for VP9 vs. 66.4% for x264) that x264 is still more data efficient than VP9.

Source:
http://en.wikipedia.org/wiki/High_Efficiency_Video...

"In a video encoder comparison released in December 2013 the HM-10.0 HEVC encoder was compared to the x264 encoder and the VP9 encoder. The x264 encoder was version r2334 and the VP9 encoder was version v1.2.0-3088-ga81bd12. The comparison used the Bjøntegaard-Delta bit-rate (BD-BR) measurement method in which negative values are how much lower the bit rate is reduced for the same PSNR and positive values are how much the bit rate is increased for the same PSNR. In the comparison the HM-10.0 HEVC encoder had the highest coding efficiency and on average to get the same objective quality the x264 encoder needed to increase the bit rate by 66.4% while the VP9 encoder needed to increase the bit rate by 79.4%"


RE: The state of VP9
By heffeque on 1/3/2014 1:41:11 PM , Rating: 2
So... quality wise, as of Jan 2014:

h.265 > h.264 > VP9

I wonder what software and hardware will be needed for h.265 playback and recording (for battery longevity).

VP9 makes no sense unless h.264 or h.265 go berserk with patent and royalty stuff.


RE: The state of VP9
By bug77 on 1/3/2014 2:36:51 PM , Rating: 2
Thanks. Dug some more into it and it seems VP9 still doesn't know what a b-frame is. It will never beat h265 w/o that, but it's surprising it can come so close to h264 as it is.


RE: The state of VP9
By sprockkets on 1/3/2014 7:28:29 PM , Rating: 2
I'm sure they know what one is - a patented technique they can't use until expired.


RE: The state of VP9
By bug77 on 1/3/2014 8:30:47 PM , Rating: 2
Crap.
Luckily, it seems it is set to expire this September. If Google was secretly working to add support for b-frames the day after this expires (and have that hw accelerated), that would be uber-cool. Wishful thinking, I know.


RE: The state of VP9
By amanojaku on 1/3/2014 1:30:46 PM , Rating: 3
quote:
Are there publicly available encoders yet? Last I checked, many features (profile 1) were still to be implemented.
I don't think so, but you can check the official page. Supposedly, it's libvpx:

http://www.webmproject.org/hardware/vp9/
quote:
Also, if anyone can link a more recent comparison to h264, I'd really appreciate it. Past comparisons I've seen were rather narrow in scope and placed VP9 either in front or behind h264.
That depends on what you mean by "in front or behind". There are at least three variables that determines the "best" codec:

1) picture quality after compression compared to before
2) bitrate for a given picture quality
3) encoding/decoding time (processor utilization)

VP9 is, in one test, at the back of the pack, with HEVC (h265) being the best in terms of bit rate reduction while maintaining quality, followed by h264. I haven't seen decoding performance numbers, but VP9 decodes slower (needs a faster processor) than h264, but faster that HVEC. You'd go with HVEC if you have the latest hardware (the developers admit they sacrificed processor utilization for picture quality and low bitrate), h264 in all other cases. VP9 currently has no advantages other than being open.
quote:
A performance comparison of H.265/MPEG-HEVC, VP9, and H.264/MPEG-AVC encoders was presented. According to the experimental results, the coding efficiency of VP9 was shown to be inferior to both H.264/MPEG-AVC and H.265/MPEG-HEVC with an average bit-rate overhead at the same objective quality of 8.4% and 79.4%, respectively. Also, it was shown that the VP9 encoding times are larger by a factor of more than 100 compared to those of the x264 encoder.
http://iphome.hhi.de/marpe/download/Performance_HE...


RE: The state of VP9
By someguy123 on 1/4/2014 7:52:05 PM , Rating: 3
I do not see the point of 4k video on youtube at all when their encoder is just absolute garbage even at the current resolutions (and for some reason much worse as you go down in resolution. Upscaling a 480 video to 1080 yields better encode quality and bitrate). 1080p on youtube is just a mess of smudged, denoised pixels, so I'm assuming "4K" will simply look slightly closer to a decent 1080p encode.


RE: The state of VP9
By bug77 on 1/5/2014 7:44:14 AM , Rating: 2
Some videos uploaded to Youtube that are marked as HD are really not. But most of them are ok, so I don't know how you came to the conclusion that their encoder sucks.
And yes, virtually everybody uploads compressed content that will be compressed again, so Youtube will never look as good as a BluRay disc. Then again, it's not meant to.


RE: The state of VP9
By someguy123 on 1/5/2014 6:58:02 PM , Rating: 2
I came to that conclusion from actually uploading videos. I've tried bruteforcing by just shoving more bitrate at it (even more than they recommended for enterprise), getting rid of reference frames, adding reference frames, increasing and decreasing encode complexity with h264, and hell even using completely lossless codecs like lagarith/huffy. End result is quality in and garbage out. Only way I've subjectively improved quality is by denoising significantly to improve compressibility, but then again the image is destroyed by denoising. I don't see the point in having a 1080p option when it looks nothing like local (or even netflix 720p) 1080.

I have yet to see a video posted on youtube that isn't heavily, significantly compressed, so I'm actually more interested in what videos you've been watching.


RE: The state of VP9
By bug77 on 1/5/2014 7:39:51 PM , Rating: 2
quote:
I'm actually more interested in what videos you've been watching


http://www.youtube.com/watch?v=P99qJGrPNLs
http://www.youtube.com/watch?v=c0i88t0Kacs
http://www.youtube.com/watch?v=OPVWy1tFXuc

I could go on, but any of the above destroys your claim that "upscaling a 480 video to 1080 yields better encode quality and bitrate". Granted, you can spot compression artifacts because, well, it's lossyly (sp?) compressed video, but since 4k video won't be shown on monitors that will be twice as large as they are today, even that won't matter as much.


RE: The state of VP9
By heffeque on 1/6/2014 1:01:33 PM , Rating: 2
The first two have no grain at all.
As for texture quality, the first two are not so bad (because of the lack of grain, the bitrate can go towards textures), but the third one... it's all washed out.
Also all suffer from terribly visible banding all around

You've just proved the above poster right and yourself wrong.

Unless you're serious about thinking that those videos are high quality. In that case, you must be used to low quality video or have a crappy TV/monitor.

Quality encoding isn't about the amount of pixels. It's about the quality of the colors, the lack of banding, the quality of the grain, the definition of the textures...


RE: The state of VP9
By bug77 on 1/6/2014 4:29:15 PM , Rating: 2
Like I said, it's compressed, so it can't be perfect. But never in a million years can you mistake those for 480p, like the guy above has claimed.


RE: The state of VP9
By someguy123 on 1/6/2014 5:48:13 PM , Rating: 2
I never claimed it was 480p. I claimed they are nowhere near a 1080p encode, which makes the distinction pointless, and I claimed upscaling from 480p to 1080p would give a better ENCODE through youtube (which is true).

The videos you posted are good examples of the horrible youtube encoder. Ironically it actually does remind me of a poor 480p upscale. Looks similar to the ps3's scaler.
http://i.imgur.com/whRw2Kg.jpg

The cyberpunk encode came out much better, but mainly due to ease of compressibility (very low motion scenes that mostly just pan). As I said before, making things more compressible improves the subjective IQ.


RE: The state of VP9
By someguy123 on 1/6/2014 6:17:41 PM , Rating: 2
So I ran a quick test to see how it would actually look and, well, the 480p upscaled actually does look a bit like the 1080p option on youtube depending on scene (though youtube doesn't seem to have dithering). The youtube encoder results are actually worse than I initially thought.


Pipes
By DaveLessnau on 1/3/2014 11:02:23 AM , Rating: 2
Interesting. More and more content is becoming available in higher and higher data-densities. More and more consumers want said content. Yet, the semi-monopolistic (or truly monopolistic) pipe purveyors (wireless and wired) keep adding data caps. Ain't life grand?




RE: Pipes
By Ticholo on 1/3/2014 11:12:31 AM , Rating: 2
They face growing irrelevance in a market where they're used to being the only game in town (often literally).
They won't go away and I hope strategic partnerships aren't the way of the future, but we still need them midterm and those practices will only keep their hostility toward their customers.
Data caps will get in the way of what their customers want and will have to go away, but unfortunately they'll just shift their energy into other forms of not providing what people want: the bandwidth they pay for, period.


RE: Pipes
By bug77 on 1/3/2014 11:13:17 AM , Rating: 2
If you ask me, that was the whole reason for data caps: add a seemingly generous limit now and profit at no additional cost when inevitably users need more bandwidth.


RE: Pipes
By lamorpa on 1/3/2014 12:28:35 PM , Rating: 2
Which carriers have added data caps recently?


VP9
By Nortel on 1/3/14, Rating: 0
RE: VP9
By Flunk on 1/3/2014 11:48:40 AM , Rating: 2
No one is dealing in uncompressed 4K video. While this is more compressed than a theatrical release even those are compressed using lossy compression. The files would simply be too large. The bit rate is about 4GB/second. That doesn't include sound and assumes 24fps, 3D would need at least double that.


RE: VP9
By Adaptation on 1/3/2014 1:15:06 PM , Rating: 2
@Flunk; You seem to be confusing uncompressed data with lossless compression. Lossless compression can be compared to a zip file, when uncompressed it is mathematically identical to the source but when encoded it is much much smaller than a bitmap stream.


What about 1440/1600p
By DanNeely on 1/3/2014 12:14:13 PM , Rating: 2
Does this mean I might be able to find a 3rd YouTube video at native width for my 30" monitor sometime this year.




RE: What about 1440/1600p
By Captain Orgazmo on 1/5/2014 9:40:43 PM , Rating: 2
TMMI
By temp01 on 1/3/2014 12:20:33 PM , Rating: 2
http://www.tmmi.us/?q=technology

Same Diff?
Better?
I believe fractal video is here in 4K.




RE: TMMI
By DanNeely on 1/3/2014 10:40:05 PM , Rating: 2
Totally different. VP9 is another conventional wavelet encoding system; the successor to Google's VP8 (which almost noone outside of Google actually used) and nominally equivalent to h.265.

AFAIK Fractal encoding's never gotten beyond the R&D stage (emphasis on research); both because the amount of computation needed to do the initial encoding is orders of magnitude higher than that of the conventional wavelet encoding that's been the basis of every mainstream codec, and because of noxious patents squatting on core technology. Even if that group is able to get reasonable performance, SoC/CPU/GPU companies would be starting from scratch in creating dedicated decoder hardware for it, and without that power consumption would be killer for anything running on battery.


"We shipped it on Saturday. Then on Sunday, we rested." -- Steve Jobs on the iPad launch

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki