Texture Analysis by Accurate Identification of a Generic Markov–Gibbs Model.

A number of applied problems are effectively solved with simple Markov-Gibbs random field (MGRF) models of spatially homogeneous or piecewise-homogeneous images provided that their identification (parameter estimation) is able to focus such a prior on a particular class of images. We propose more accurate analytical potential estimates for a generic MGRF with multiple pairwise pixel interaction and use them for structural analysis and synthesis of stochastic and periodic image textures. [ABSTRACT FROM AUTHOR]