Estimation in pickard random fields and application to image processing

Abstract

The goal of this study is to present a method of noisy image restoration and segmentation. This algorithm, of E-M type, is at the limit of two widely used techniques: general Gibbs models and models born from signal processing. It is based on the Pickard random fields which allow full advantage of the bidimensional structure of the image to be taken, giving a better exploitation of the interaction of rows and columns in the image.