Comparison between universal data-compression algorithms applied to Markov sources

A comparison between universal data compression algorithms applied to binary Markov sources is presented. Four algorithms are considered (1) The Ziv-Lempel (Z-L) algorithm proposed in 1978. (2) A state conditioned version of the Z-L algorithm. (3) Rissanen's algorithm (1986). (4) A sequential a...

Full description

Saved in:
Bibliographic Details
Published in:IEEE Seventeenth Conference of Electrical and Electronics Engineers in Israel, 1991 pp. 111 - 114
Main Authors: Amir, I., Plotnik, E., Weinberger, M.J., Tal, N.
Format: Conference Proceeding
Language:English
Published: IEEE 1991
Subjects:
Algorithm design and analysis
Analytical models
Arithmetic
Data compression
Decoding
Entropy
Probability
ISBN:9780879426781, 0879426780
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A comparison between universal data compression algorithms applied to binary Markov sources is presented. Four algorithms are considered (1) The Ziv-Lempel (Z-L) algorithm proposed in 1978. (2) A state conditioned version of the Z-L algorithm. (3) Rissanen's algorithm (1986). (4) A sequential algorithm recently presented by Weinberger, Lempel and Ziv (see IEEE Trans. on Information Theory), that performs an online estimation of the source structure and uses an arithmetic code. It is shown via simulation that, actually, the redundancy of the Ziv-Lempel (Z-L) algorithm is considerably smaller than the analytical upper bounds given in (2). Also, the average codeword length in the state conditioned version is, in most cases, smaller than in the original Z-L algorithm, even when conditioning with respect to a set of states that differs from the actual set of states of the source. The redundancy for the sequential algorithm of (4) is almost the same as that for Rissanen's code.< >
ISBN:9780879426781
0879426780
DOI:10.1109/EEIS.1991.217740