hufflen.m

This file contains a new and improved version of the Huffman coder, (June 29. 2001). The name is Huf

function HL = HuffLen(S)
% HuffLen     Find the lengths of the Huffman code words
% Based on probability (or number of occurences) of each symbol 
% the length for the Huffman codewords are calculated.
% 
% HL = hufflen(S);
% ------------------------------------------------------------------
% Arguments:
%  S  a vector with number of occurences or probability of each symbol
%     Only positive elements of S are used, zero (or negative)
%     elements get length 0.
%  HL length (bits) for the codeword for each symbol 
% ------------------------------------------------------------------
% Example:
% hufflen([1,0,4,2,0,1])  =>  ans = [3,0,1,2,0,3]
% hufflen([10,40,20,10])  =>  ans = [3,1,2,3]

%----------------------------------------------------------------------
% Copyright (c) 1999.  Karl Skretting.  All rights reserved.
% Hogskolen in Stavanger (Stavanger University), Signal Processing Group
% Mail:  [e-mail]   Homepage:  http://www.ux.his.no/~karlsk/
% 
% HISTORY:
% Ver. 1.0  28.08.98  KS: Function made as part of Signal Compression Project 98
% Ver. 1.1  25.12.98  English version of program
% Ver. 1.2  28.07.99  Problem when length(S)==1 was corrected
% Ver. 1.3  22.06.00  KS: Some more exceptions handled
%----------------------------------------------------------------------

if nargin<1
   error('HuffLen: see help.')
end
% some checks and exceptions
if (length(S)==0)          % ver 1.2
   warning('HuffLen: Symbol sequence is empty.');   % a warning is appropriate
   HL=0;
   return;
end
I=find(S<0);
S(I)=0;
if (sum(S)==max(S))
   disp('HuffLen: Only one symbol.');   % a message is appropriate
   HL=zeros(size(S));       % no Huffman code is needed
   return;
end

%  Algorithm "explained" in Norwegian:
%   En bygger opp "treet" ved