On Bayesian segmentation of zone of interests of scalar scenes

We consider a problem on the segmentation of zones of interest which represent scene fragments, containing a given object together with its certain neighborhood. We state the problem as that of the Bayesian classification of pixels of a zone of interest into two classes. We prove that for the estimation of unknown a priori probabilities and conditional distributions of classes it is sufficient to know the area of the object and to have an image of the zone of interest. The results of the segmentation of zones of interest in model and real scenes by the proposed method are offered.