Guest post by The Free Software Foundation announced today that it will be supporting the WebM project. As Brett Smith, license compliance engineer for the FSF, said in a press release: “Google has been working to bring together a broad coalition of organizations to support WebM, which will go far to help it become the video codec of choice with HTML5. e want the world to know that we also support WebM: with its developer-friendly patent license and free software reference implementation, it’s a good choice to help ensure the Web fulfills its promise of providing a free way for the world to communicate.”
I wonder if the FSF has done any secret research into the patent aspect of WebM. If any patents against it were revealed, I would expect that the FSF would be sort of forced to revoke their support for the format. I doubt they would support WebM if they felt that it was at all risky. So I feel safer now.
There have been no patent claims made against WebM to date.
When Google purchased On2 the process took a long time, over six months I believe, because Google undertook an extensive and thorough patent search around the VP8 codecs. I don’t believe that FSF have done any research, but Google certainly did do the research as part of due diligence over the acquisition of On2. According to Google, their search uncovered no patent problem with VP8.
Beyond the fact that both VP8 and H264 use the same basic approach to video compression (which was discovered about sixty years ago with the introduction of colour TV), the VP8 codec exhibits fundamentally different behaviour to the H.264 codec. VP8 produces blur in parts of the video that are moving quickly, whereas H.264 does not blur the picture but rather adds “artefacts” or extra little shapes and features which are not real. This different behaviour points to an underlying fundamental difference in the compression methods used for the chroma information (which is the part of the video information which one can afford to be lossy).
There is very little reason to think that WebM infringes on any H.264 patents, in fact it is just about the same level of suspicion that H.264 might infringe upon On2 patents (now owned by Google) for VP8.
Edited 2011-01-20 00:23 UTC
Discovery of colour TV has nothing to do with the changes between VP8 and h.264. The colour TV has only introduced a ingenious concept how to separate colour information (chroma) and luminosity information which is why the b&w TV still worked even if the signal was sending colour picture, colour TV could just add colour information when available. This separation of colour and luminosity information has been helpful in video compression as you just could compress colour information more because the eye is less sensitive for distortions in colour than it is to luminosity. Not only that – the chroma information is transformed to 1/4 of its original resolution in the format that is most used for video compression today – YUV 4:2:0. This however has nothing to do with the difference between h.264 and VP8 as they use the same format or even many other formats as well.
Blurring is also no indication of “fundamental differences in compression methods” as encoders “love” to blur when they use PSNR as a quality measuring as PSNR gives a higher grade for blurred picture. Blurring is the indicator that VP8 does not yet include any psycho-visual optimizations. For example this optimizations are what has been added to Theora ptalarbvorm to produce a lot better (and less blurry) picture. Look at the first picture (or download the video) in this introduction to ptalarbvorm optimizations: http://people.xiph.org/~xiphmont/demo/theora/demo9.html. If you read carefully the a lot better video encoded with ptalarbvorm has actually produced a worse PSNR value than the thusnelda encoded video.
Psychovisual optimizations will eventually be added to VP8 (probably similar as they have been added to Theora) and then it will also produce (visually) a lot better quality video that now – and the gap between h264 will be even smaller.
h264 and VP8 are actually very similar in techniques used, but still intentionally different enough that VP8 don’t violate h264 patents (or VP8 doesn’t use the technique that could not be safely work-around).
Up to this point, we agree almost exactly. Right up to when when you say “has been helpful in video compression as you just could compress colour information more because the eye is less sensitive” we agree exactly. The eye has far lower colour resolution is the way I would put it, but your words I think are saying the same thing. The term is luma, BTW, and not luminosity as these mean slightly different things.
http://en.wikipedia.org/wiki/Video_compression
Agree with that as well. The reaon why I mention it is that this concept, where the required resolution of the colour (chroma) part of the image is a lot lower, is a 60-year-old “invention” originally discovered in the development of colour TV, and virtually all video codecs use this as the primary area for reduction of the information required to be encode. Because this technology is so old, it is not patentable now.
In other words, this is NOT an area of difference between H.264 and VP8, it is in fact an area of commonality. This is what I said. We agree here, too.
My actual point was that, although VP8 and H264 use common techniques here, this technique is not patentable. Both codecs are perfectly at liberty to use this particular technique, and both of them do. Patents are not at issue here, though.
Here, we disagree. Blurring in VP8 is very probably due to averaging of the chroma signal in the time domain (i.e. over a few frames), in order to further reduce the resolution of chroma data to be encoded. Thus, if most of a scene is still except for, say, an actors rapid arm movement in thowing a punch, if you freeze the frame mid-punch in VP8 you will see the arm kind-of smeared across the picture occupying the areas where it was for a few frames in the original. If you look at the same arm in the same frame of the same video encoded in H.264, you will see the same arm in a single, definite place, but distorted in its shape. The arm might have a very block fist on the end, for example.
This is direct evidence a fundamental difference between the compression methods. I think the VP8 codec reduces chroma resolution to be encoded via motion blur, not image blur, and h.264 has motion pixelation. H.264 appears to have reduced the resolution of encoded chroma information by averaging it over a number of adjacent pixels.
This has nothing to do with “psychovisual optimizations”, this is an apparent fundamental difference in techniques used.
http://en.wikipedia.org/wiki/Video_compression#Intraframe_versus_in…
I think they are even more diffrent from one another than you apprently surmise.
Edited 2011-01-20 11:21 UTC
That’s huge for Google.
I think we can now take the way of HTML-5 & videos.
As an IT professionals and enthusiasts, I can now seriously think about developing web sites with HTML-5 video tags.
It is quite more safer, lesser risk to dump all the work done for a patent question.
Now even in a bad case of patent lawsuit, I think the community will be large enough to find an alternative solution for WebM.