Dirac video codec 1.0 released

> Dirac is a royalty-free video codec ...

Your point is valid.  I have changed the title and text to at least give readers some idea
that it is a video codec.  I need to be more careful about context when putting those items
together, thanks!

jake

How hard did you look?  A google search of "dirac sample video" returned a pointer to the
sourceforge page with sample videos:

http://sourceforge.net/project/showfiles.php?group_id=102...

Hi

I tested it yesterday in Mandriva 2008. 
Hopefully these "Aunt Tillie" steps are useful also for others.

1) Install gstreamer0.10-devel 
    - urpmi libgstreamer0.10-devel
2) download, build and install the latest liboil source package
     - ./autogen.sh
     - make & make install
3) download, build and install the latest Schrödinger dirac decocer
   with gstreamer plugin
     - ./autogen.sh
     - make (if gstreamer0.10 was recognized it will build gstreamer plug-in under gst
directory)
     - make install
--> You should have:
lamikr@pia gstreamer-0.10]$ ls -la /usr/lib/gstreamer-0.10/libgstschro.*
-rwxr-xr-x 1 root root   1047 2008-03-10 18:51 /usr/lib/gstreamer-0.10/libgstschro.la*
-rwxr-xr-x 1 root root 264989 2008-03-10 18:51 /usr/lib/gstreamer-0.10/libgstschro.so*

4) Get some dirac file 
- transfered video clip from my videocamera over firewire with Kino to DV file
- converted DV file to mp4 file with Kino (oggconvert did not handle DV files directly)
- searched oggconvert-0.3.1.tar.gz from internet (python gui frontend app)
- extracted oggconvert-0.3.1.tar.gz (can be launched from any directory)
- selected mp4 file for the source in oggconvert
- selected Dirac as a target format
- selected output file name and pressed convert
--> dirac file with ogv extension was created

5) Play the dirac file with gstreamer plug-in installed in step (3)
- gst-launch-0.10 playbin uri=file:///home/lamikr/kino/swamps_on_lake_in_winter.ogv
--> video with picture opened

Mika

Software patents aren't recognized or legal in Europe. My guess is that the BBC is on very
safe grounds - in Europe - but that Dirac has potential problems in the USA. (Even if the
patents have expired, your average website owner or distro developer can't afford the lawyers.
If the BBC muscled in to back Dirac users, or growled loud enough, that might help, but I
don't recall seeing them do that. If anything, I seem to recall the BBC being rather timid
when being directly hastled for putting their own concerts online.)

We are often asked about patents and, in particular, about arithmetic coding patents.

Clearly we are aware of patents surrounding arithmetic coding and have been very careful to
avoid them. We have conducted a literature search and base our arithmetic coder on prior art
literature dating back to Shannon’s 1948 “A Mathematical Theory of Communications” and more
directly to Rissanen & Langdon’s 1979 “Arithmetic Coding” in the IBM Journal of Research &
Development as well as other literature.

Patents in the US (prior to 1995) expire 17 years after grant. If there are any basic patents
on arithmetic coding they have already expired or will do so soon. But, actually, we don’t
think there are patents on the basic concept of arithmetic coding. The patents that do exist
are on practical implementations and optimisations.

Most arithmetic coding patents relate to ways of avoiding multiplies and the ways of gathering
symbol statistics in an efficient way that allows the use of multiplier less implementations.
Bear in mind that these patents were taken out in the late 1970’s and early 1980’s when
avoiding multipliers was a big deal.

We have taken the basic concepts and produced our own version of arithmetic coding. Our
version doesn’t seek to avoid the use of multiplication. This is because there is no need with
modern CPUs and modern (FPGA) hardware to do so, Using multipliers no longer incurs a
performance penalty. That is something that has changed in the more than 20 years since
arithmetic coding emerged.

Since we don’t seek to avoid multiplies in our arithmetic coding the claims on optimised
implementations do not apply. Of course there may be those that think they have patents on the
basic concepts. But, if challenged, we can produce prior art dating back many years that would
prove such patents to be invalid.

Even though we haven’t sought to optimise our arithmetic coder by eliminating multiplies we
have, nevertheless, done considerable optimisation. The basic arithmetic coder is very simple
and produces results that are comparable to those obtained by the arithmetic coder used in
H264. We also have the advantage that arithmetic coding in Dirac is easy to parallelise, in
contrast to H264. This allows us the potential for considerable software optimisation and also
allows us to achieve higher bit rates (useful for professional applications).

We could have patented our arithmetic coder. We chose not to do so because to patent it would
have been contrary to the basic principles of the Dirac project. Our arithmetic coder is now
published in our source code and so can not, now, be patented.

We are confident that the intellectual property in embodied in Dirac, including that in
arithmetic coding, does not require a patent licence, in Europe, the US or elsewhere.

Schroedinger is the high-performance implementation of the Dirac codec. 
AFAIK Dirac also has codec software, but this isn't focussed on 
performance, but just to demonstrate how it works.

Alex

See here:
http://dirac.sourceforge.net/

Quote:
The purpose of the Dirac codebase has changed over time. It is no longer intended as the basis
of a real-time practical implementation: for that you want Schrodinger.
End of quote

The Schro coder is an optimised version of the Dirac codec. It is an independent
implementation, not a fork. The development has been supported by the BBC who have been
working with Fluendo to produce Schro. We are currently working to ensure that both Schro and
Dirac are interoperable and conform to the specification. Schro and Dirac are currently
interoperable and we code video using Dirac and play back with the Schro decoder.

Schro is an implementation of Dirac.

Both codec implementations have their own advantages and disadvanatges, particularly the
coders. The Dirac coder, currently, has the edge on the Schro coder in terms of compression
performance, but the Schro coder is faster. We will be working on the Schro coder to improve
its compression performance over the next few months. Sometime later in the year I expect the
Schro coder to equal the compression performance of the Dirac coder.

We intend to maintain the Dirac codebase, at least for the time being. I expect another
release of Dirac (0.9.2) in a few weeks and this will not be the last. Dirac has an advantage
as an experimental test bed. Improvements incorporated in Dirac will probably later be
integrated with Schro. Nevertheless Dirac and Schro will remain interoperable.