The development of Markov random field theory and applications on image segmentation algorithm

Abstract

Image segmentation algorithm is to divide the images into several regions with specific and unique characteristics, and is an important technology to extract the interested target. Image segmentation is the key step to realize the research from general image processing into image analysis, and is vital preprocessing method of image recognition and computer vision. We cannot obtain correct recognition if we do not have correct segmentation. Nevertheless, the only basis of segmentation process is brightness or color of pixels in an image. In the processing of computer automatic segmentation, we experience several problems, such as uneven illumination, effect of noise, indistinct part in image, and shadow, and these factors may cause false segmentation. In order to overcome the disadvantages of the traditional segmentation algorithm, in this paper, we propose a novel segmentation algorithm based on Markov Random Field. The segmentation algorithm proposed in this paper is based on Markov Random Field Mode and Bayesian theory, and we determine the objective function in image segmentation problem on the basis of optimality criterion of statistical decision and estimation theory. Some optimization algorithms are used to obtain the maximum possible distribution of Markov Random Field which satisfy these conditions. The experimental result reflects the effectiveness and robustness of our algorithm. As a supplement, we analyze the development trend of the Markov Random Field theory.