SAR Image Segmentation with Variable Classes Using RJMCMC Algorithm

In SAR image segmentation, automatically determining the number of classes is a critical and difficult problem. To this end, this paper presents a statistics based SAR image segmentation approach which can automatically determine the number of classes and segment the image simultaneously. First of all, a given SAR image is modeled on the assumption that intensities of its pixels satisfy identical and independent Gamma distributions. The Bayesian paradigm is followed to build image segmentation model. Then a RJMCMC (Reversible Jump Markov Chain Monte Carlo) scheme is utilized to govern the segmentation model, which determines the number of classes and segments the image. In the proposed RJMCMC algorithm, four move types are designed, including splitting or merging real classes, updating parameter vector, updating label field, birth or death of an empty class. In order to verify the proposed algorithm, testing is carried out with real and simulated SAR images, respectively, and the results show that the proposed algorithm works well and efficient.