Contrast enhancement of badly illuminated images based on Gibbs distribution and random walk model

Abstract

In the paper a new approach to the problem of contrast enhancement of grey scale images is presented. The described method is based on a model, which treats the image as a toroidal, two dimensional lattice, the points of which possess a potential energy. On in this way defined lattice, a regular Markov chain of the positions of this particle can be investigated. The probability of a transition of the virtual particle from a fixed lattice point to a point belonging to its neighbourhood can be determined using the Gibbs canonical distribution, defined on an eight-connectiviy system. The idea of the presented algorithm consists in determining the stationary probability vector of the Markow chain. The new algorithm registers the visits of the wandering particle and then determines their relative frequencies. It performs especially well in case of images with nonuniform brightness.