Skip to Main Content
Proposed a novel image segmentation method based on Markov random field (MRF) and context information. The method introduces the relationships of observed image intensities and distance between pixels to the traditional neighborhood potential function, so that to describe the probability of pixels being classified into one class. We transform the segmentation process to maximum a posteriori (MAP) by Bayes theorem. Finally, the iterative conditional model (ICM) is used to solve the MAP problem. In the experiments, this method is compared with traditional expectation-maximization (EM) and MRF image segmentation techniques using synthetic and real images. The experiment results and SNR-CCR histogram show that the algorithm proposed is more effective for noisy image segmentation.