Huffman Coding and Arithmetic Coding

Functions that compress integer sequences.

Description

This file contains MATLAB functions, m-files, that do Huffman coding and arithmetic coding of integer (symbol) sequences. Complete coding may be done by calling an easy to use main program (or main function), where input argument is the sequences you want to compress and the output is the compressed bitstream, as a vector of bytes. Decoding is done just by switching the arguments. These main programs are: the Huffman coder, Huff06 an updated version of the Huff03 function sbmitted earlier, and two versions of the arithmetic coder, Arith06.m and Arith07.m. The zipped file also contains examples and a contents file.

Example of use:
 y=Huff06(xC); % encoding
 xC=Huff06(y); % decoding
where xC is a cell array of column vectors of integers representing the symbol sequences and y is a column vector of non-negative itegers (bytes) representing the code, 0 <= y(i) <= 255.

For more information see the web page: http://www.ux.his.no/~karlsk/proj99/

• MATLAB release: MATLAB 6.1 (R12.1)

[Norbert Marwan]

I'm rather frustrated with this programme.

First, the names of the functions contain uppercase letters. However, sometimes these functions are called with lowercase letters (eg in Huff06 the HuffLen), which will lead to an error on case sensitive systems. It is much better to use only lower case letters!

The encoder/decoder is very slow. Maybe some vectorization of the loops can be applied in order to speed up the algorithm.

The help and Contents are not correctly formatted, eg the mentioned functions in the Contents cannot be used as links. An workung example, how the encoder/decoder works, should be given in the helptext.

[Antti Stenvall]

I think that this isn't working properly.
I have a sequence (propabilities for symbols)
p = [49 26 12 4 4 3 2];
And HuffLen+HuffCode gives me coding:
  Symbol 1 gets code: 0
  Symbol 2 gets code: 10
  Symbol 3 gets code: 110
  Symbol 4 gets code: 11100
  Symbol 5 gets code: 11101
  Symbol 6 gets code: 11110
  Symbol 7 gets code: 11111

The First Three rounds are okay, but then it makes wrong encoding. It just switches somehow the symbols. Or am I just stupid?

[Arsalan Naeem]

need to improve more, but overall good work

[asem eshtyeh]

please can you send to me code in matlab abuot 2.4Kbps linear predcition coding
  i will very thanks

[derasriya sunil]

 

[Yeditepe University]

poor programming and documentation.

[Enrique Estalayo]

Poor documentation and non-clear programming algorithms. Difficult to follow the structure of the code

[jeya shree]

nice

[M Khan]

I got the following error message when I tried to execute TestHuff.m
-----------------------
BitsUsed=2090 bits=2109
??? ｴﾗｰ: ==> Huff06>EncodeVector
Huff06-EncodeVector: Logical error, (BitsUsed~=bits).

[Matlab User]

I also get errors just by running TestHuff:

BitsUsed=2090 bits=2109
??? Error using ==> Huff06>EncodeVector
Huff06-EncodeVector: Logical error, (BitsUsed~=bits).

Error in ==> Huff06 at 140
      [bits, ent]=EncodeVector(x); % store the (abs and/or log) values

Error in ==> TestHuff at 38
[y, Res]=Huff06(xC, Level, Speed);

Using Matlab 7 R14 SP3

[hassan abid]

 also get errors just by running TestHuff:

BitsUsed=2090 bits=2109
??? Error using ==> Huff06>EncodeVector
Huff06-EncodeVector: Logical error, (BitsUsed~=bits).

Error in ==> Huff06 at 140
[bits, ent]=EncodeVector(x); % store the (abs and/or log) values

Error in ==> TestHuff at 38
[y, Res]=Huff06(xC, Level, Speed);

Using Matlab 7 R14 SP3
reply as soon as possible

[Matlab User]

"coding of integer (symbol) sequences", but double data is used?! -> Just theoretical value.

[Burak Olcay]

 

[Karl Skretting]

Authors remark May 2011: An updated version is available from web-page: http://www.ux.uis.no/~karlsk/proj99/index.html (comp.zip). Some few errors has been corrected the last 10 years, but the program is essential like before.
My e-mail is now: [e-mail]