Research on Medical Image Segmentation Method Based on Markov Random Field Model
|Course||Signal and Information Processing|
|Keywords||image segmentation Markov Random Field model chaotic simulated annealing algorithm class-adaptive|
Medical image segmentation is a fundamental technique of medical image processing. In order to distinguish normal tissue and abnormal pathological changes, we need to segment the medical image.This paper first gave a survey on image segmentation methods. Among those methods, Markov Random Field (MRF) show many advantages on medical image segmentation, so the author focused his research interest on this method. Then, the basic theory about Markov random field model was introduced.Secondly, after carefully analyzed the advantages and disadvantages of MRF-based image segmentation method, the author proposed an improved MRF model, which integrates region, priori knowledge and boundary information of the image. The proposed model incorporates geometry shape boundary information, and improves the objective function of traditional MRF model. In order to improve the speed of segmentation, Chaotic Simulated Annealing (CSA) algorithm was introduced to solve the improved MRF model for the first time. CSA algorithm can greatly enhance the speed of global optimization.In addition, the author developed an interactive MRF Image Segmentation Demo software for medical image. It was used for processing and displaying medical image. The method was validated on both phantom and clinical images. Experiments on clinical cardiac MR images showed that the improved MRF model has high performance on segmenting medical images. The experimental results illustrated that this model was strong anti-noise, accurate and efficient, especially for the weak boundary and concave region.Finally, considering homogenous and isotropic MRF in the interior of the region while inhomogeneous and anisotropic MRF on the boundary, an adaptive coupling coefficient was proposed. A segmentation method based on MRF model with class-adaptive coupling coefficient was presented. The proposed segmentation method was proved effective for medical image and it was a self-adaptive segmentation method with more exquisite result.