Object recognition using Markov spatial processes

Abstract

The Bayesian approach to image processing based on Markov random fields is adapted to image analysis problems such as object recognition and edge detection. In this context the prior models are Markov point processes and random object patterns from stochastic geometry. The authors develop analogues of J. Besag's algorithm (1986). The erosion operator of mathematical morphology turns out to be a maximum likelihood estimator for a simple noise model. The authors show that the Hough transform can be interpreted as a likelihood ratio test statistic. >