Image Segmentation Using Energy Minimization and Markov Random Fields

Image segmentation is one of the hot fields of computer vision. In this paper, we propose a novel Markov random fields image segmentation algorithm. According to Gibbs distribution and MRF equivalence, image segmentation problem is transformed to minimize the posterior energy function corresponding to the labeling problem. The energy function can be efficiently minimized using the expansion move algorithm which is one of the most effective algorithms in graph cuts. The data term parameter estimation method using an iterative process is similar to the EM (expectation maximization) algorithm. Experimental results are provided to illustrate the satisfactory performance of our method on both synthetic and remote sensing images.