Introduction
Previous: Contents
Next: Overall architecture
Dirac is an experimental Open Source video codec, initially
developed by BBC Research and Development. Its design is that
of a conventional hybrid motion-compensated codec, except that
the the function normally performed by a block transform in
most MPEG standard and proprietary codecs is instead performed
by the wavelet transform.
Dirac performs well in comparison with other codecs, and
certainly much better than MPEG-2 (probably about twice as well).
However, performance was not the only factor driving its design.
Dirac is intended to be simple, powerful and modular, both
conceptually and in implementation.The aim is to produce a codec
which replaces the enormous variety of tools normally required
for high-performance video compression with a much simpler
design, with far fewer choices for the encoder to make.
Ultimately, we want Dirac to develop into software that can
be used for real-time encoding and decoding in image-processing
applications: it is not supposed to be an algorithmic
demonstrator. The software is a long way off this point at
present, but with time we hope to get there.
We also want a wide variety of developers to be able to
work with Dirac, and so we have tried to document Dirac as
completely as possible. In addition to this algorithm document,
we have provided doxygen-based comments on the reference code itself.
We have produced a specification of the Dirac bit stream and decoding
operations to allow others to develop compliant coders and decoders using
as much or as little of the original Dirac software as they please. The
high-speed Schrodinger implementation has been developed entirely from
the specification.
If you want to use the Dirac reference code, you are referred to
the
README
file in the distribution bundle for a description
of what the various options mean.
Previous: Contents
Next: Overall architecture
|