Dirac: BBC Open Source Video Codec 523
NickFitz writes "Need To Know this week has a piece about Dirac, a BBC R&D project to produce a video codec, which has been released as an Open Source project. From BBCi: 'Dirac is a general-purpose video codec aimed at resolutions from QCIF (180x144) to HDTV (1920x1080) progressive or interlaced... Our algorithm seems to give a two-fold reduction in bit rate over MPEG-2 for high definition video (e.g. 1920x1080 pixels), its original target application. It has been further developed to optimise it for internet streaming resolutions.'"
can I get a nice quality of 1 hour video on a CD? (Score:3, Interesting)
if this can get me to 700 MBs at the same quality (about 85 in the 3vix) that would rock!!!
Re:Another one? (Score:2, Interesting)
WOOO... NO MORE REAL PLAYER!! (Score:3, Interesting)
BBC Archive (Score:3, Interesting)
Here's hoping it leads to more videos (Score:5, Interesting)
Their documentaries are so interesting that I often choose to watch them over other movies or shows I may have on my computer. Bravo BBC.
Re:WOOO... NO MORE REAL PLAYER!! (Score:5, Interesting)
Though FYI this was developed by their R&D team, which is, according to a bloke behind Dirac I spoke to at the recent Linux Expo in London, quite separate from other parts of BBC tech.
I asked him about their recent OGG trials, and he said that not only did he know nothing about it, his dept. didn't even play any part or have any influence. Crazy big corporations!
Re:Duplicating work? (Score:4, Interesting)
Patent free (Score:5, Interesting)
Very cool. But then all the engineers in their R&D dept. are apparently very anti software patents, and have been doing their bit writing letters to that effect
BBC Archive (Score:5, Interesting)
Did you ever consider that they intend to use this for their plans to put their archives online? [bbc.co.uk]
The poster stated that "It has been further developed to optimise it for internet streaming resolutions" which is one way for such a thing to be distributed. Have a look here. [theregister.co.uk] The register states that "The BBC's new media director, Ashley Highfield, said that a P2P network will allow the BBC to handle the volume of traffic it expects when the Internet Media Player (IMP) goes live. The IMP will enable users to download or stream content to their PC, laptop or palmtop computer."
If this is the case then Aunty Beeb is well underway to providing the tools we will need for accessing their archives.
-- Enditallnow
Great news (Score:3, Interesting)
I am glad that the BBC has recognised the need for this codec to be Open Source. It means that everyone, not just those beholden to private corporations, will get the chance to experience BBC content. The BBC is also a highly influential body; I would be surprised if other European content providers did not display an interest in this. PAL was a joint development between the BBC and its German counterpart; SUSE is German.
This is going to be one to watch.
Here's the True Reason. (Score:3, Interesting)
How do they do it all for free? (Score:5, Interesting)
My real question to Brits here is: How well is this burden accepted by the British people? Are the BBC TV and radio stations in the UK really non-commercial? I know the US government gives money to PBS and NPR but I don't know how it compares (especially per capita) to what the British government must spend on the BBC. It must cost a fortune and they are effectively supplying (IMHO) a good quality product for free to the rest of the world via internet and shortwave. I imagine some of the international motivation of the BBC stems back to the days of the old empire. It almost seems too good to be true.
I assume that NPR and others like it around the world pay to carry the program. Maybe that earns a lot.
This question came up in my mind the other day when the wi-fi radio was mentoned here on
Re:seems they were ahead of me (Score:2, Interesting)
Gotta love Auntie (Score:4, Interesting)
-psy
There's two kinds... (Score:4, Interesting)
a) 640x360 vid at 1x bandwidth
b) 640x360 vid at 9x bandwidth
c) 1920x1080 vid at 9x bandwidth
a) and c), MPEG4 will win. b) will be much much closer. What you define as "low" bitrates really depends on resolution. The dual-layer DVDs coming now should be able to do full HDTV resolution with somewhat better quality than a 1CD DVDrip. Since 2CD rips typically use 3-400mb on AC3 track, actually not that far away from 2CD rip. But something like 8-10Mbit (aren't they usually 1Mbit today?) is hardly a low-bitrate stream in my opinion...
Kjella
Let's be clear. (Score:2, Interesting)
For one, MXF and AFF are two wrappers for convential video codecs, which add invaluable meta data to the file.
Dirac stands out as being a unique codec as it concentrates wavlets, motion compensation and arithmetic coding. This is not your standard codec.
Comment removed (Score:2, Interesting)
The BBC patented the codec (Score:5, Interesting)
They also said that while they had no objection to paying licensing fee's per se, and that they did pay MS and Real, these were so inflexible in their licencing that scaling up operations was problematic. Their expressed hope was that with such a codec widely adopted they could massively scale up operations such as streaming without being crippled with licencing costs, or having the administrative burden of unwieldy licensing schemes.
Background (Score:5, Interesting)
A matter of disclaimer: I've done some work on Dirac, for BBC, over the last several months. Here's a bit of background on Dirac:
By nature, Dirac has many similarities to existing algorithms such and MPEG-2 and H.264/AVC -- however, Dirac is an original invention that uses wavelet transforms, arithmetic encoding, rate-distortion optimization, variable block-size motion compensation, and hierarchical motion estimation in some new and unique ways. Again, this is a research project, so there's much experimentation to be done!
As a research project, Dirac continues to be analyzed, optimized, and documented. What you're seeing now is very preliminary code; I suspect it will improve and evolve dramatically in the coming months, both in terms of clarity and functionality. The goal is to produce a universal codec, which is one reason behind the open source move.
The codec source code is licensed under dual MPL/GPL licenses.
Dirac is modular, and thus well-suited to implementation with an object-oriented programming language. The reference engine is written in ISO Standard C++, and has been tested under various forms of 32- and 64-bit Linux, as well as under Windows 2000/XP.
I'll try to answer questions here, to the best of my ability.
Is this named after the same Dirac? (Score:2, Interesting)
We have a library named after him here at Florida State University. Is this the same guy?
Re:Duplicating work? (Score:3, Interesting)
The block artifacts in DCT/FFT based approaches such as MPEG-2 are due to the fact that video has local detail (well, duh!) but you're using a globally repeating waveform (FFT = sine, or DCT = cosine) to encode it. In order to represent local detail with a DCT-based tranforrm you have to divide the image into blocks to localize the transform, and when you throw away the lower order transform coefficients (which is basically what transform-based encoding is all about) those block boundaries become apparent - block artifacts.
In contract, wavelets are spatially localized / non-repeating, so you can represent local detail without needing to introduce blocks. But, when you throw away your lower order coefficients you will see bits of the image represented by only the higher order wavelets, which (wavelets being smooth) means that there'll be blurriness.
The BBC codec might still choose to use blocks for motion compression (I don't know), but the above basically explains the difference between the two, and definitely applies to the difference between image compression using DCT (JPEG) vs wavelet (JPEG2000) compression.
Re:Background (Score:3, Interesting)
Those very issues are currently under consideration. I'd be interested in hearing opinions on the matter.
Re:seems they were ahead of me (Score:2, Interesting)
And that duty's not just 'something they should do to be nice', it's the law -- the Royal Charter and Agreement which the BBC operates under makes this clear.
Incidentally, the BBC are pretty big Linux/F/OSS proponents. Their entire BBC Interactive digital service is run from Linux systems. There are some very clever folk at the Beeb, used to 'rolling their own' technologies. I'd imagine something like Linux is the perfect tool to do this kinda stuff with. They've got vast amounts of in-house software and hardware which they've developed over the best part of a century.
Anyone else remember the BBC Micro? They, and the Open University have to take quite a significant amount of credit for kick-starting interest in computing in the UK during the late 70's and early 80's.
Incidentally, the Open University, [open.ac.uk] whose course materials BBC 2 broadcast nationally and for free, is a fantastic organisation. Like the BBC, it was founded by Royal Charter, and is another example of how something spawned by government need not be tethered to it. It's been a highly respected institution from which to graduate for decades. Here's some info [open.ac.uk]
Re:Background (Score:5, Interesting)
Real-time encoding is one goal, but probably not our first target. I believe it will be possible, especially with parallelisation for multi-processor workstations.
As for patents: I am not a lawyer, nor do I play one on Slashdot. I do not speak for BBC in any way; I'm just answering questions because everyone in Britain is home eating dinner at the moment.
It's a very tricky world out there right now. Arithmetic coding can be implemented without hitting patents, I believe -- and the modular design of Dirac should allow a different coding scheme -- say, Huffman -- to be implemented if patents become an issue.
Re:UK to US shows and back again (Score:2, Interesting)
Not really. Friends is about... well, friends,and Coupleing is about coupling. Yes, the format has obvious similarities, but the the characters are different, the type of humour is different, the relationships are all different.
There was a made-in-the-US version of "Red Dwarf"? Did not know that.
They tried twice. Two pilots. Actually, according to one of the writers the second pilot was great until the studios got their hands on it.
Here are two extremely successful UK to US sitcom adaptations:
I was not aware of those. I stand corrected. Perhaps it's just a recent thing that they've been oversanitising them, or something.
Re:The BBC patented the codec (Score:4, Interesting)
If they patented the codec but then release an implementation under the GPL, it follows that they are effectively granting anyone permission to, as patent lawyers say, "practice the patent" as long as they do so in a GPL program that is a derivative work of what they released. It would be good of them to say so explicitly, though.
There are a growing number of GPL-licensed patents now (patents where the owner permits GPL implementations). IBM has done this with a number of their patents.
Re:The BBC patented the codec (Score:3, Interesting)
In broad terms, Yes. If you establish the invention as prior art no-one can get a subsequent patent for that exact invention. However there are important technical catches in patent application procedure in relation to timing of the disclosure compared to any patent application; so it may not work in some circumstances.
Also, someone may patent an improvement in your invention: it would be unpleasant if the BBC gave Microsoft a head start - however that would also be so for an offensive patent or by an examination of any non patented Free software.
Counting arithmetic operations is irrelevant (Score:1, Interesting)
That was on display at the linux user show. (Score:3, Interesting)
James
Dirac welcome to work with 100% open Helix Player (Score:2, Interesting)
Secondly, this summer we will be releasing the RealPlayer 10 for Linux, which is built on top the Helix Player and includes the non-open source components of MP3, MPEG4, Flash, and RealAudio and RealVideo.
Both the Helix Player and RealPlayer 10 have a Mozilla plug-in and are going to be a major no cost update for your current RealPlayer 8.