Dirac: BBC Open Source Video Codec

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.'"

Comments

Duplicating work?

[bobbis.u]

Seems like a bit of a waste of license payers money when there already a several open source video codecs (eg. Ogg Theora, 3ivx). What does this offer that those don't?

Re:Duplicating work?

[breakinbearx]

It seems as if their codec performs very well (losslessly?) at very high resolutions, i.e. the ones the Beeb would need for HDTV. This would provide better (visual) quality programs over existing lines e.g. cable, satellite. I have a feeling the codec will not have a practical use for everyday computer users, but will be used more for video production.

Re:Duplicating work?

[Anonymous Coward]

Codecs have varying quality/compression tradeoffs for different source material, intellectual property hassles (not all which are known up front), varying API support, etc. So the more that are available the better.

Re:Duplicating work?

[dubiousdave]

I have a feeling the codec will not have a practical use for everyday computer users

I don't know about that. I'd be rather pleased if MythTV could record twice the HD content on the same hard drive space on my computer, or, for that matter, if TiVo were to use it for the same purpose.

Re:Duplicating work?

[evilviper]

there already a several open source video codecs (eg. Ogg Theora, 3ivx)

With Xvid, ffmpeg/libavcodec, and any others based on MPEG-4, the code may be open source, but you can't use it legally, without paying for an MPEG-4 license. MPEG-4 is a lot like MP3 in that sense.

Theora would be nice, but it's perpetually Alpha... I was excited about it at first too, but now it seems it's going to take another year before the code is even in beta, and probably two years before it reaches 1.0, when there will be ports to non-Linux platforms. By then, it will be about as advanced as MPEG-1 is today... Way behind the times.

However, VP3 (the codec Theora is based upon) is a rather good codec (despite the brain-dead review it got at doom9). It is free, open source, etc. There are encoders and decoders for Windows/Mac, and numerous decoders for Unix systems. It would really work great, and I have no idea why it hasn't been more popular to date.

If there was some program that could encode VP3 video on Unix systems, I would be using VP3/Vorbis excluively for encoding everything. However, avifile, MPlayer, ffmpeg, none can encode to VP3, so it seems Unix systems are out of luck.

That said, I'd bet the BBC will be doing their encoding on Windows or Mac OS machines anyhow, so I don't know why they don't use VP3.

Re:Duplicating work?

[Doomdark]

Don't be an ass. What he is saying that degradation is less irritating: just like in general analog mobile phone signal sounds "better", even with sub-optimal physical quality, than digital signal; this because artifacts sound so unlike 'normal' degradation. Even if physically measured, square distance (or whatever measure is usually used) from signal might be the same (or even 'better' for digital transmission)

To put it another way: same amount of physical distortion corresponds to different amounts of subjective distortion. In general, "natural" distortion is more pleasing to human eye and ear (well, brains, eventually) than "non-natural" ones. And blocky MPEG artifacts are worse than wavelet-generated non-symmetric degradation.

Re:Duplicating work?

[JimDabell]

That's irrelevent, as porting Theora to Windows is almost certainly much easier than coming up with a whole new codec.

Re:Duplicating work?

[Uzik2]

There are a LOT of projects out there that
never generate decent product, or even any
product at all. They can't wait for some
other group with no stake in their future
to write something they need.

From my reading of Theora, it's just the
streaming container, not a codec. Vp3 is
the codec. I don't know if it's suitable
for what they want to do with it.

They may also have felt they could do better.

Re:Another one?

[dubiousdave]

Our algorithm seems to give a two-fold reduction in bit rate over MPEG-2 for high definition video (e.g. 1920x1080 pixels)

That seems to answer your question, even without reading the article.

it's open source!

[g00bd0g]

Because it's open source and being developed/supported by one of the bigger distibuters of video content on the web!

Re:Another one?

[Anonymous Coward]

Yes, why would anyone need a new, free high quality codec that is backed by a huge media company? Why not use one of the patented, commercial codecs that Microsoft or Real.com want us to use?

Re:WOOO... NO MORE REAL PLAYER!!

[WormholeFiend]

and with a bit more luck, the CBC (Canadian Broadcasting Corporation) [cbc.ca] will also ditch RealPlayer as well.

A bit wary

[bsd4me]

I would be a bit wary of a codec that claims to be all things to all people, ie supporting broadcast-quality HDTV and internet-quality video

Video codecs typically have ``sweet-spots'' for resolution and bitrate. The MPEG specs work well for higher bitrate video, and we have several codecs that work well for lower bitrate video.

Also, MPEG video quality can vary from encoder to encoder. The specs only define the bitstream, and the encoder can do what it wants. This is why there is a huge difference between the quality that Media Cleaner produces versus a multi-chip hardware encoder found in a cable plant.

MPEG4?

[cybermace5]

Why would they compare it to MPEG2? In order to impress me, you'll have to compare quality and bitrates with MPEG4.

Re:MPEG4?

[Jameth]

MPEG4 is not that special at very high bitrates. MPEG4 is for low bitrates almost exclusively. This makes for small files which look good enough, but not files which look perfect.

Fantastic News

[stephenry]

This is absolutely brillant news!

Ever since I heard that the BBC plans to put their achive on the internet it was clear that they would be far better served developing their own video codec. As a British Citizen, I am glad that those who have paid television licenses do not have to pay an additional toll in the form of Real Player.

My question is...

[LilMikey]

How are they going to convince set-top manufacturers to support their codec or conglomerates to broadcast it? It's already been proven a hundred times over the superior and/or open rarely win out to their more profitable brethren. All the article states is there's a 'hint of a chance' of it being adopted by big media...

Re:My question is...

[spotteddog]

MONEY

They claim to have a process that can cram twice the information in the same bandwidth. If I am broadcasting HDTV (MPEG-2 at 19Mb/s) and I switch, I now have two channels available.

If I'm a sat. TV company (DirecTV, Dish, etc.) I have doubled my channel capacity without launching any expensive new sats.

Given that it is free, I'd say they have a good chance of having major players adopt their codec in the near future.

Duplicating work? Not really.

[Ride-My-Rocket]

Seems like a bit of a waste of license payers money when there already a several open source video codecs (eg. Ogg Theora, 3ivx). What does this offer that those don't?

Does the current work being done on Linux seem like a waste of time and money, when there are already several other operating systems (Windows, Macintosh, Unix) available? And don't try and use the argument "but those are closed source; open source is better!" argument -- in the end, it's just software people use, regardless of the licensing / development model.

Getting back on-topic: apparently it offers the BBC something that warrants the time, effort and money required to fund such an undertaking. At the very least, it's yet another example of big companies using open-source to reduce costs and/or fulfill their own specific needs, and can only encourage other companies to fund future OSS development efforts.

Re:BBC = british government

[Neophytus]

Since the BBC is the media-organ branch of the British government, this means government-owned codecs. Is this a good thing?

Unlike some of the stateside media organisations the BBC is actually one of the world's most impartial media organisations. I'm not saying they are perfect but some US news bulletins I cann't watch without laughing.

Re:great now when I download a fansub

[Anonymous Coward]

None of those are codecs. XVideo is a video conferencing program, you're thinking of the XviD codec. MKV is the Matroska video container, and OGM is the Ogg Media container.

And quit your bitching.

Re:can I get a nice quality of 1 hour video on a C

[Lord Kano]

In all seriousness, you must be doing something wrong.

I can get two hours of good quality video onto a CD with DivX.

LK

source code quality

[hak1du]

Perhaps even more impressive than the improved bit rates is that the source code actually looks competently written and is small. It also seems to use C++ in a reasonable way: to achieve just around the right amount of abstraction, without building a useless, general framework.

Re:source code quality

[ChaoticCoyote]

The framework is changing as we profile and analyze the code.

Speaking for myself (independent of Dirac), the best C++ code is the simplest code. Just because a feature exists doesn't mean it must be used -- and conversely, just because a feature can be used poorly doesn't mean it should never be used. The goal is to use the right C++ features for the job, and avoid become lost in a nest of complex classes and templates.

There's nothing about Dirac, BTW, that requires C++, or even object-oriented programming.