Hi Tim et-al,
Would it be possible for you to give us a progress update on where things are at with Dirac?
things have been quiet since April/May and i know you were talking about interesting events ahead with Dirac Pro at the time.
This is doubly intriguing for us closet Dirac enthusiasts given the iplayer business, so how about it? :)
Good idea - yes we've been a bit remiss about updating people here.
For those that don't know, Dirac Pro is the I-picture only subset of Dirac. We've expanded the Dirac spec to include very low delay coding by re-arranging the wavelet coefficients, as this is really useful in broadcast production. So Dirac Pro is either whole-picture coding (like JPEG 2000) or very low delay coding.
Dirac Pro is being proposed as an international standard through the SMPTE, where it will become (if ratified) VC-2 - Microsoft's WM9 is VC-1. So we've been writing two specs - the full Dirac spec, which is on our website, and a cut-down version omitting motion compensation for SMPTE. We also have to write reference software (which we'll also release open source), a "Recommended Practice" (RP) standard and provide reference test bitstreams for SMPTE - so there is a whole shed-load of work there!
Our aim, if VC-2 works and gets adoption in the industry, is probably to come back and complete the job on the full codec in a subsequent standard, but we're feeling our way here.
We exhibited at the International Broadcast Convention (IBC) again this year, with our hardware partners Numedia. Numedia have produced a number of hardware DiracPro/VC-2 codecs for broadcast applications, like carrying HD video over SD (SD-SDI) links with very low delay, and multiplexing HD signals together over single links and so on. So you can now buy Dirac HW :-)
All this has diverted our energies from the core software codec, but Dave Schleef has been doing a great job on Schrodinger, which is an optimised version of Dirac. He's had to track the many versions of the specification as it's evolved. Schro is pretty near completion, and is probably more feature-complete than Dirac itself now. We're going to do another release of Dirac which incorporates recent improvements in compression and also supports interlace coding (i.e. coding fields instead of frames).
Looking at the codebases though, Schrodinger is in much better shape than Dirac, so what we'd like to do is make sure that the compression performance of Schro matches (or exceeds) Dirac and then freeze development on the Dirac codebase altogether and go forward with Schro instead. The Dirac code was trying to do too many different things and grew in a haphazard way.
I've probably forgotten loads of stuff, but I hope this helps.
Many thanks for the update.
So will there be a v1.0 release of Schro in 2007, or at least a public release which is usable?
If so, do you imagine adoption and use will be muted while people wait to see if the final Dirac/Schro breaks compatibility?