Research On Level Sets Segmentation Method With Local Clustering And Global Intensity Properties

With increasing use of Computed topography(CT)and Magnetic resonance(MR)images for diagnosis,treatment planning and clinical studies,it has become almost compulsory to use computers to assist radiological experts in clinical diagnosis,treatment planning.Reliable algorithms are required for the delineation of anatomical structures and other regions of interest(ROI).Image segmentation based on level set model is a kind of dynamic contour segmentation methods which are widely applied to medical image segmentation.The Advantages of level set method are mainly summarized as practical,convenient,fast and robust.In the latest research results among the level set method has been developed tono need to re-initialize,and joined the confinement energy parameters.In this paper,we are aiming to find precise medical image segmentation under gray variation.We have proved the advantages of our model by stereo analysis:1)Analysis of the intensity inhomogeneity for image segmentation algorithm advantages and disadvantages of the widely used today.The basic idea of some of the current algorithms are based on improved energy terms,the energy functional containing global and local information as possible,but the fact that the two can not have both,the paper generalized Gaussian kernel introduced energy functional,before the image segmentation,estimate the shape parameter generalized Gaussian functions,the global information capture,the energy functional building process to get rid of the constraints of the global energy terms,strengthening the algorithm for image segmentation ability gray variation.2)With a new way to obtain local information:local property based K-means clustering algorithm.Local information acquisition method proposed in this paper,first define a local gray clustering function,for each point are partial classification.Thus the greatest degree of local gray collected information.3)From the two-phase multiphase segmentation to multiphase segmentation,step by step description of the algorithm.Given a number of division parameter update model to improve the algorithm for image segmentation multi-target capability.