A fast algorithm of image segmentation based on Markov random field
Zhihui Li, Meng Zhang, Haibo Liu · 2012
Markov random field (MRF) is a common used consistent approach in image segmentation. However it has a drawback of low computation speed. A fast definite algorithm of Markov random field is proposed in this paper, which is based on common used frame of maximum posterior probability and Potts model. A representation of binary label field is adopted to describe the membership of each pixel to one of classes. The label fields are derived by iterations of computation. During one time of iteration the label fields are updated on basis of the results of last iteration. The energy function of MRF is computed by mean filter. The running speed of the algorithm is increased while the smoothness effect remains as same as other algorithms. The experiment results in the paper show the effectiveness of the algorithm.