Texture Segmentation Based on Multiscale Fuzzy Markov Random Field Model in Wavelet Domain

Since the inadequate use of statistical information in the multiscale Markov random field model in the wavelet domain, the fuzzy theory is introduced into the model and a new multiscale fuzzy Markov random field model in the wavelet domain is presented. Firstly, the model which is based on the fuzzy theory defines a fuzzy probability field, which is used to describe the degree of every pixel belong to which segmentation region in each scale in wavelet domain. During evaluating parameters stage of the following feature field, the whole pixels' features in the same scale are taken into account. Then, the indicator field which reflects the energy of every pixel that belong to certain segmentation region is induced from the feature field. Finally, a three-step interaction iterative segmentation steps based on the Bayes rule is induced. The experiments compared with the ICM and MRMRF algorithms have proved the proposed method’s availability