Huffman Coding routines, Re: Octave-Forge, Source coding: Tunstall code, Huffman code

[Muthiah Annamalai]

Hello there,
I added the Huffman coding routines in Octave-forge, on the lines of
the 'other-leading-brand' with 2 features that allow for variations
in code generation including minimum variation option.

3 functions:
1. huffmandict,
http://octave.cvs.sourceforge.net/octave/octave-forge/main/comm/inst/huffmandict.m?revision=1.1&view=markup
2. huffmanenco,
http://octave.cvs.sourceforge.net/octave/octave-forge/main/comm/inst/huffmanenco.m?revision=1.1&view=markup
3. huffmandeco,
http://octave.cvs.sourceforge.net/octave/octave-forge/main/comm/inst/huffmandeco.m?revision=1.1&view=markup

You can check the code in the latest Octave-forge CVS, and if possible
do some testing. AFAIK its all ok, but the speed is not really nice.
Especially the dict and deco methods, if someone has time and energy
to read and code the O(N*log(N)) dict & deco routines, it will be just
amazing. Till then we all need to live in O(k.N^2) and O(N^3) times ;-(

For someone who has the time useful pointers are: 
http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?isnumber=30738&arnumber=1423737&count=169&index=103
and http://www.cs.auckland.ac.nz/software/AlgAnim/huffman.html

Cheers
Muthu



On Sat, 2006-09-30 at 20:22 +0200, David Bateman wrote:
> Muthiah Annamalai wrote:
> > Hi David,
> > I have some source coding algorithms, I would like to see included into 
> > Octave-forge communications package, as I see it already has
> > implementations for arithmetic encoding & decoding,  some quantization
> > algorithms etc.
> > 
> > Im presently adding it with the Information theory package, but with
> > the other codecs in comm' package we might have to start a new
> > branch of source coding algorithms only or merge with existing.
> > Merging seems less work, and useful.
> > 
> > -Muthu
> 
> Agreed, but please look at the
> 
> 
>  arithenco
>  arithdeco
>  dpcmdeco
>  dpcmenco
>  dpcmopt
> 
> in matalb and see if there are ways of making any code compatible with
> them. I developed my code against R13, and so the R2006b comms toolbox
> might have slightly different functions and interface so check the
> mathworks web site as well..
> 
> D.