Maximum Likelihood decoding of binary images corrupted by Ising-like noise

CESAR F. CAIAFA

Trieste, ITALY

Workshop; Common Concepts in Statistical Physics and Computer Science; 2007

International Centre forTheoretical Physics - ICTP

 In this work, the decoding of binary images corrupted by additive noise is analyzed. A noise process based on the two dimensional Ising model is proposed which allow to model the spatial interaction among pixels usually found in real images. The Maximum Likelihood (ML) decoding algorithm is derived and an approximation based on the Bragg-Williams formula is used to construct a even simpler decoder. The efficiencies of both methods showed to be equivalent for a wide range of the model parameters. Some illustrative examples of decoding are presented.