On2 Releases VP6 video codec

A reader writes:"On2 Technologies, the folks who brought you the open-source VP3 video codec (now managed by Theora.org), have released our latest codec, VP6. Highlights include hi-def support with no encoder restrictions, real-time encoding at full D1 resolution, and substantial performance & quality improvements over VP5. Best of all: no "patent pooling" restrictions or external licensing fees, a la MPEG-4."

Comments

This ad paid for by On2

[Anonymous Coward]

On2 Technologies, the folks who brought you the open-source VP3 video codec (now managed by Theora.org), have released our latest codec, VP6.

Glad to see people can just submit press releases.

Licensing

[Wesley Felter]

Sure, VP6 has simpler licensing because it is completely proprietary, but H.264 is supposedly patent-free and it has the advantage of being a published standard with mulitple competing implementations.

Re:Hmm

[TrekkieGod]

many people don't even know or want to know how to install XviD

Right up to the point where they download their first XviD video file. Then, like anyone else, they'll pay the price and install the darn codec (after posting on countless message boards asking why the heck their video won't play).

It's a necessity thing. Nobody wants to install what they don't need. However, the people who *make* videos experiment all the time to get the best quality they can...and once that happens, the others are forced to follow. Heck, if I can be made to install RealOne on my computer, anybody can be made to install anything. It was the hardest compromise of my principles I've ever made.

Does anyone really care?

[Datasage]

I dont, i'll continue to use Divx or Xvid. Being that both are either free or really cheap (for the pro ver of divx).

I dont buy thier claims that its any better than divx. To start with, thier samples are biased and most likely have been changed to show an advantage in thier product.

If they do in fact have a better quality then the comperable codecs, is it worth the cost of licening this codec for just a slight improvement in quality?

And xine does ogg theora now ...

[timothy]

related note; the older On2 codec which has become the basis of Ogg Theora isn't dead: Ogg Theora is listed as supported in the newest version of Xine, as mentioned yesterday.

timothy

3D ideas in movie codecs?

[TeknoHog]

AFAIK, many video codecs use some kind of FFT/DCT/wavelet compression for individual frames, and then some rather different techniques to account for the similarities/changes between subsequent frames. As a physicist I find this rather too complicated; it would be more natural to treat time as a third dimension, and just use FFT/DCT/wavelets or whatever for the whole 3D package (probably split into NxNxN cubes).

Quick googling shows that such codecs have in fact been investigated. But are these ideas used in any current codecs?

Of course every lossy codec has its problems, but there's at least one reason why I'm intrigued by these 3D style codecs: it could be easier to balance resources between single-frame precision vs. motion. With fast action we're less sensitive to details, so the codec could use more information for motion and less for the image details. And vice versa for slow scenes. This would happen automatically if the codec always preserved the strongest percentage of frequency/wavelet components (which is a standard way of signal de-noising).

Re:3D ideas in movie codecs?

[plastik55]

Well, such a scheme might produce somewhat better compression (especially since the impulse responses of visual neurons resemble 3d wavelets, naturally suggesting a 3d scheme for efficient representation.) I once implemented a scheme using 3d wavelets for optic flow detection, with some pretty good results on canned data.

The drawback is that it requires more complicated encoders and decoders, which use more memory and more memory bandwidth due to the need to decode several frames at once. Today's hardware isn't quite up to the task, especially for encoding in realtime or decoding on a set-top box.

IMO, 3d transforms might help boost efficiency but they probably won't lead to a more mathematically "elegant" codec. After all, the task is to produce an image that fools the human visual system into percieving the image as real thing, using the smallest number of bits in the encoding. Since the problem definition is based on something as big and complicated and poorly understood as the human visual system, the problem solution is going to reflect that complexity. The more advanced video codecs get, the more they will reflect the complexities of the visual system. Look at mp3--it's basically a pretty simple quantized DCT scheme for audio, but getting good performance out of it requires the encoder to know about and exploit all sorts of psychoacoustic masking effects, basically embedding a model of the auditory system into the encoder.

Re:Hmm

[TrekkieGod]

The difference between DivX files and XviD files is the FourCC ("Four Character Code" or something). DivX and XviD are just differernt encoders, not different formats.

True enough, although certain features of mpeg4 encoding can be, at any one time, implemented in one, and not the other. I haven't been keeping up with the progress of either xvid or divx lately, but I know that at one point xvid was ahead in B-frames support, which caused some compatibility problems between it and divx.

Of course, all that aside, even if they are completely compatible with each other, if you're talking about a person who doesn't know how to install XviD, do you really want to recommend to that person that he should try changing the FourCC code? I assume not, since you later said:

That's why I think, the people who encoded the video file should change the FourCC to DIVX, try to decode the file using DivX, and (if it works without problems) release the file with the DIVX FourCC.

Which, I guess if you test before-hand, would work just fine. Still, I don't know if I agree with your manner of thinking, and many encoders might disagree also. Why wouldn't you ask instead that divx encoders change the FourCC to XVID? Because divx is more popular? Well, that's a great disservice to the xvid developers, who could use the extra help in making *their* codec more popular by causing more people to download it. After all, I personally prefer the open source xvid to divx...The divx 5 team actually tries to charge you for the "pro" version of their codec which has enhanced features for encoding.

Re:Install ffdshow and other video tools mentioned

[Joe Tie.]

I'd love to see a write up on video encoding on linux, maybe I'll do one myself.

I forgot to mention, that if you do take this on, Doom9's Linux forum is probaly the single best source of information out there to start with.

I think there's a real need for a site centered on video editing, capturing and encoding under Linux. It was probaly the single biggest challenge for me moving to Linux, simply because all the information out there is so scattered. Compiling some information is something I keep meaning to do, but never seem to be able to find the time and energy for.