Non-Proper Variable-to-Fixed Length Arithmetic Coding

Suk-hee CHO  Ryuji KOHNO  Ji-hwan PARK  

IEICE TRANSACTIONS on Fundamentals of Electronics, Communications and Computer Sciences   Vol.E81-A   No.8   pp.1739-1747
Publication Date: 1998/08/25
Online ISSN: 
Print ISSN: 0916-8508
Type of Manuscript: PAPER
Category: Information Theory and Coding Theory
VF code,  VF arithmetic coding,  non-proper code,  data compression,  

Full Text: PDF(703KB)>>
Buy this Article

The VF (Variable-to-Fixed length) arithmetic coding method combines the advantage of an ordinary stream arithmetic code with the simplicity of a block code. One of the advantages of VF codes is that the transmission errors or channel errors do not propagate infinitely and are restricted to the block in question. In this paper, we propose a modified type of non-proper VF arithmetic coding method that defines an input alphabet subset according to both the number of codewords in the current codeword set and input symbol probability and that splits the codeword set completely for a newly defined alphabet subset when the codeword set becomes smaller by each splitting. The proposed coding method carrys out independence of each codeword and guarantees that there is no collision while there is a waste of codeword(s) in conventional AB-coding due to collision. We examine the performance of the proposed method and compare it with that of other VF codes in terms of compression ratio and algorithmic complexity.