Quantization of both stationary and nonstationary Gaussian sources with Voronoi constellations

Author

Pépin, Christine ; Belfiore, Jean-Claude ; Boutros, Joseph

Author_Institution

Dept. of Commun., Ecole Nat. Superieure des Telecommun., Paris, France

fYear

1997

fDate

29 Jun-4 Jul 1997

Firstpage

59

Abstract

This paper is concerned with the use of Voronoi constellations for vector quantization of stationary and nonstationary Gaussian sources. First, stationary Gaussian sources are treated. The mean-squared error (MSE) criterion is used as the distortion measure and approximations for the granular and the overload MSE are derived. Then, nonstationary Gaussian sources are introduced. A new technique for quantization using a weighted metric and a high-diversity Voronoi constellation as the codebook is presented. Each part provides simulation results for Voronoi constellations based on D₄, E₈, and A₁₆

Keywords

Gaussian processes; image coding; least mean squares methods; rate distortion theory; transform coding; vector quantisation; MSE criterion; approximations; codebook; distortion measure; granular MSE; high-diversity Voronoi constellation; mean-squared error; nonstationary Gaussian sources; overload MSE; simulation results; stationary Gaussian sources; transform encoded image; vector quantization; weighted metric; Constellation diagram; Encoding; Laplace equations; Large-scale systems; Lattices; Noise reduction; Probability; Rate-distortion; Statistics; Vector quantization;

fLanguage

English

Publisher

ieee

Conference_Titel

Information Theory. 1997. Proceedings., 1997 IEEE International Symposium on

Conference_Location

Ulm

Print_ISBN

0-7803-3956-8

Type

conf

DOI

10.1109/ISIT.1997.612974

Filename

612974