Markovian Fields as Limits of Markov Chains

摘要

Markovian fields are useful stochastic models for statistical treatment ofdigitized images, among other applications. In this paper we present some simulation results about the convergence properties of one of the available methods to simulate Markovian fields: the Gibbs sampler. We use the observed values of the mean magnetization, the pseudo-likehood estimates, and the observed short and long range correlations to study the problems associated with the stopping time of the algorithm for different values of the parameter.