The first in-depth technical analysis of VP8
Even shorter summary of VP8: meh.
Yeah, but it's still a good thing since it might put some pressure on MPEG LA not to get too greedy.
And also, if there is going to be a video codec patent war, I'm happy for Google to be in the fight and not just Mozilla or Theora.
Does this Slashdot comment make sense Mark? Most video codec patents revolve around implementations of the discrete cosine transform, Huffman coding, chroma sub-sampling, and bilinear interpolation. All of these techniques are older than the patent examiners who approved the patents and indeed the judges who will try the cases. It's all mathematics, every last bit. These patents are all essentially equivalent to patenting the tetrahedron.
There is nothing the USPTO will not give a patent for. As such, there is absolutely nothing in the universe past or present which can be declared patent free wherever the authority of the USPTO is recognised. Maybe a suit against VP8 could expose (and fix?) some of this craziness? That's probably hopeful, but who knows...
As far as VP8 goes, Microsoft has already said they will only support it (in Windows 7) as a user installed downloadable codec. In other words, you won't be able to rely on it for Windows. And Apple has so far been silent, but I would be surprised if they support it by default either at this point. So it's sort of dead in the water from that standpoint at the moment (but might still be very helpful as outlined above.)
On the plus side, hardware companies are on board, including AMD, ARM, NVIDIA, and Qualcomm (but noticeably not Intel). Plus, "Broadcom announced today that its motion video acceleration solution for mobile phones, VideoCore, should gain WebM support by Q3 2010." So the hardware support (especially for mobile devices) that I suspected would never materialize for Theora is clearly going to be there for VP8.
So even though it does seem a little "meh" right now, it's looking way better than theora (just the hardware support alone is a huge win) and maybe in a couple generations it will become more competitive - hopefully by 2016 when the free royalty period for H.264 expires.
Not really important, but sort of funny in a snarky way: John Gruber points out that Mozilla (or at least Mozilla evangelist Christopher Blizzard) is now saying that The VP8 codec represents a vast improvement in quality-per-bit over Theora and is comparable in quality to H.264. Wasn't Mozilla previously arguing that Theora was as good as H.264? What happened?
I thought Mozilla were the good guys! You write like you hate them!
You take everything I write so extremely. I like Mozilla. I think FireFox is a great product and I use it all the time (FireBug is indispensable for web development.) But that doesn't mean I have to agree with every one of their stances. And one stance I didn't agree with was their claim that Theora produced as good quality video as H.264. I didn't think this was the case, and I think they claimed it out of an overly ideological belief that was clouding their scientific judgement. And now apparently Mozilla (or one of their evangelists at least) is admitting as much. But like I clearly said, this is just a snarky point that isn't of any importance to the major debate. Mozilla and Opera and Google are all supporting VP8 in their browsers and that's a great win for the web in general. Microsoft is not, and I suspect Apple will not as well. They are the bad guys here (with the caveat that we'll see what Apple announces).
Not sure how you got from there to me "hating" Mozilla.
I was noting "way better than Theora" plus "evangelist" plus asking if they were contradicting themselves. I said "like you hate them," not
that you hate them. "Overly ideological belief clouding their scientific judgement" is pretty damning.
I raise this to get a sense of whether Mozilla is going over to its own version of the dark side. I realize even "open" programs can be taken over by idiots. Some recent examinations of Wikipedia have shown me how nerds who are wrong can take over the conversation just through sheer persistence.
Theoria is much, much worse than h.264. Anyone saying otherwise is wrong. Jim is spot on noting the shifting story.
From what little I know, VP8 is very roughly comparable to h.264 baseline profile, which is inferior to h.264 main profile and high profile.
H.264 is old at this point, so VP8 is old.
VP8 apparently steps on some patents. Yeah, it's all math. But there are some innovations in the math that increase compression efficiency.
The slashdot comment is partly right, partly wrong, partly an oversimplification. Huffman coding is almost 60 years old. Transforms go back at least to Fourier and Gauss (200 years). However, B-frames are new. Weighted prediction is new. Arithmetic Binary Coding (used in CABAC) is new. The transform used in h.264 is an approximation of the discrete cosine transform that has the properties of no rounding errors and no multiplies -- that's new and very cool.
It is true that the USPTO is way too liberal with patents. They aren't crazy liberal; I've had things kicked back.
Patenting novel algorithms isn't the same as patenting the tetrahedron. The "DCT" in h.264 is a good example. Sure, transforms are old. Discrete transforms are old. Fast algorithms for discrete transforms are old. But the 264 transform is novel, non-obvious (until you've seen it), and better.
From the linked analysis, VP8 steps on patents, just the way VC1 did. What that means is to be determined. VP8 could easily become another VC1 (i.e. minor player due to the fact that it's a slightly inferior, highly derivative codec).
Mark, your thoughts on the patent situation are what I feared - that it's more complex then just that the USPTO is always wrong! Seems like a pretty big mess but maybe there is no clean way to run it.
Tom: "[W]ay better than Theora" is just the facts. I haven't seen this disputed anywhere. And I call Christopher Blizzard a Mozilla "evangelist" because that is his official title at Mozilla and I wanted to make it clear that as far as I know this is just one non-engineering person at Mozilla saying this and not necessarily the party line from the company at large. It does appear to contradict what Mozilla had been saying in terms of the H.264 vs. Theora debate where they were, I think wrongly, claiming that Theora was "as good" in quality to H.264. But I in no way think that makes Mozilla evil or marks any turn to the darkside.
But like you say, I didn't interpret your "like you hate them" quite correctly so the end of my reply was probably a little too strong.
It's likely VP8 will be a player. A star or a cameo?
|
Even shorter summary of VP8: meh.
- mark 5-19-2010 8:37 pm
Yeah, but it's still a good thing since it might put some pressure on MPEG LA not to get too greedy.
Maybe a suit against VP8 could expose (and fix?) some of this craziness? That's probably hopeful, but who knows...And also, if there is going to be a video codec patent war, I'm happy for Google to be in the fight and not just Mozilla or Theora.
Does this Slashdot comment make sense Mark?
As far as VP8 goes, Microsoft has already said they will only support it (in Windows 7) as a user installed downloadable codec. In other words, you won't be able to rely on it for Windows. And Apple has so far been silent, but I would be surprised if they support it by default either at this point. So it's sort of dead in the water from that standpoint at the moment (but might still be very helpful as outlined above.)
On the plus side, hardware companies are on board, including AMD, ARM, NVIDIA, and Qualcomm (but noticeably not Intel). Plus, "Broadcom announced today that its motion video acceleration solution for mobile phones, VideoCore, should gain WebM support by Q3 2010." So the hardware support (especially for mobile devices) that I suspected would never materialize for Theora is clearly going to be there for VP8.
So even though it does seem a little "meh" right now, it's looking way better than theora (just the hardware support alone is a huge win) and maybe in a couple generations it will become more competitive - hopefully by 2016 when the free royalty period for H.264 expires.
- jim 5-20-2010 3:09 pm
Not really important, but sort of funny in a snarky way: John Gruber points out that Mozilla (or at least Mozilla evangelist Christopher Blizzard) is now saying that
Wasn't Mozilla previously arguing that Theora was as good as H.264? What happened?- jim 5-20-2010 3:14 pm
I thought Mozilla were the good guys! You write like you hate them!
- tom moody 5-20-2010 3:40 pm
You take everything I write so extremely. I like Mozilla. I think FireFox is a great product and I use it all the time (FireBug is indispensable for web development.) But that doesn't mean I have to agree with every one of their stances. And one stance I didn't agree with was their claim that Theora produced as good quality video as H.264. I didn't think this was the case, and I think they claimed it out of an overly ideological belief that was clouding their scientific judgement. And now apparently Mozilla (or one of their evangelists at least) is admitting as much. But like I clearly said, this is just a snarky point that isn't of any importance to the major debate. Mozilla and Opera and Google are all supporting VP8 in their browsers and that's a great win for the web in general. Microsoft is not, and I suspect Apple will not as well. They are the bad guys here (with the caveat that we'll see what Apple announces).
Not sure how you got from there to me "hating" Mozilla.
- jim 5-20-2010 3:46 pm
I was noting "way better than Theora" plus "evangelist" plus asking if they were contradicting themselves. I said "like you hate them," not that you hate them. "Overly ideological belief clouding their scientific judgement" is pretty damning.
- tom moody 5-20-2010 6:02 pm
I raise this to get a sense of whether Mozilla is going over to its own version of the dark side. I realize even "open" programs can be taken over by idiots. Some recent examinations of Wikipedia have shown me how nerds who are wrong can take over the conversation just through sheer persistence.
- tom moody 5-20-2010 6:21 pm
Theoria is much, much worse than h.264. Anyone saying otherwise is wrong. Jim is spot on noting the shifting story.
From what little I know, VP8 is very roughly comparable to h.264 baseline profile, which is inferior to h.264 main profile and high profile.
H.264 is old at this point, so VP8 is old.
VP8 apparently steps on some patents. Yeah, it's all math. But there are some innovations in the math that increase compression efficiency.
The slashdot comment is partly right, partly wrong, partly an oversimplification. Huffman coding is almost 60 years old. Transforms go back at least to Fourier and Gauss (200 years). However, B-frames are new. Weighted prediction is new. Arithmetic Binary Coding (used in CABAC) is new. The transform used in h.264 is an approximation of the discrete cosine transform that has the properties of no rounding errors and no multiplies -- that's new and very cool.
It is true that the USPTO is way too liberal with patents. They aren't crazy liberal; I've had things kicked back.
Patenting novel algorithms isn't the same as patenting the tetrahedron. The "DCT" in h.264 is a good example. Sure, transforms are old. Discrete transforms are old. Fast algorithms for discrete transforms are old. But the 264 transform is novel, non-obvious (until you've seen it), and better.
From the linked analysis, VP8 steps on patents, just the way VC1 did. What that means is to be determined. VP8 could easily become another VC1 (i.e. minor player due to the fact that it's a slightly inferior, highly derivative codec).
- mark 5-20-2010 7:00 pm
Mark, your thoughts on the patent situation are what I feared - that it's more complex then just that the USPTO is always wrong! Seems like a pretty big mess but maybe there is no clean way to run it.
Tom: "[W]ay better than Theora" is just the facts. I haven't seen this disputed anywhere. And I call Christopher Blizzard a Mozilla "evangelist" because that is his official title at Mozilla and I wanted to make it clear that as far as I know this is just one non-engineering person at Mozilla saying this and not necessarily the party line from the company at large. It does appear to contradict what Mozilla had been saying in terms of the H.264 vs. Theora debate where they were, I think wrongly, claiming that Theora was "as good" in quality to H.264. But I in no way think that makes Mozilla evil or marks any turn to the darkside.
But like you say, I didn't interpret your "like you hate them" quite correctly so the end of my reply was probably a little too strong.
- jim 5-20-2010 7:10 pm
It's likely VP8 will be a player. A star or a cameo?
- mark 5-20-2010 7:59 pm