A Stochastic dynamical system for edge detection

Abstract

A new stochastic algorithm for edge detection has been presented. This algorithm is based on time-homogeneous Markov chains with the same kind of transition probabilities as in Monte Carlo algorithms and is inherently sequential in nature. The study includes a survey of classical edge detection techniques and stochastic techniques based on simulated annealing. A parallel algorithm has also been developed and has been shown to be asymptotically equivalent to the sequential algorithm. Experiments which demonstrate the efficacy of these algorithms for edge detection of two-dimensional images (in the absence as well as presence of additive noise) are presented.