All papers are zipped postscript.
PPMZ has long been the best algorithm for lossless compression. I publish here an analysis of the algorithm and why it works so well.
The PPMZ paper (42k) , "Solving the Problems of Context Modeling". Published here in March, 1998. (described work finished in 1995-6 ; I have written this at the request of some friends).
Submission to DCC 96 :
Several new techniques are presented on high order context modeling, low order context modeling, and order-0 arithmetic coding. Emphasis is placed on economy of memory and speed. Performance is found to be significantly better than previous methods. This paper was accepted for the poster section of DCC96 (meaning that only a 1 page abstract will be published).
The following papers are not finished-quality papers & have not been submitted for publication.
A new model to replace bit-stuffing or bit-plus-follow is presented, using a lossless method called the Virtual Queue. Note added : the Virtual Queue is now the standard method of arithmetic coding, used in the MACM coder, and all newer "byte-wise" arithmetic coders.
A new statistical coder, similar to Shannon-Fano coding, is presented which is extremely general, and may be implemented to any efficiency. This coder joins only 3 other known statistical coders : Shannon-Fano, Huffman, and arithmetic. This coder is interesting primarily only for theoretical reasons, not for its efficiency. Note added : BRE is a nice way to unify all the statistical encoders and make the power of the arithmetic coder clearer.
Charles Bloom / cb at my domain
The free web counter says you are visitor number