Abstract:
It is well known that stochastic models, such as MRF and AR model,
suffer from their poor capability to describe complicated textures. By
introducing filters into these random field models appears to be a new
approach to overcome the problem. In this paper, we present a new model
for texture analysis, which integrates filter theory into random field
based on Maximum Entropy theory. And parameter estimation employing
Markov Chain Monte Carlo (MCMC) is proposed for this new model.