Home – Home – Jornals – Avtometriya 2005 number 4

A method for gray-scale image restoration is proposed. It is supposed that the image is linearly blurred and observed in additive Gaussian noise. The image is considered as a realization of a Gibbs random field described by the Gauss-Markov autoregressive model. Increasing restoration quality is attained by introducing to the Gibbs model a hidden level containing borders between areas with a relatively slowly changing brightness. Results of applying the approach to real image recovering are presented.