Research Of Level Set Models For Image Segmentation And Their Applications On Medical Image Segmentation

In the development of the active contour model,the level set(LS)model has been heavily promoted by the emergence of research in the field of image segmentation.Level set model combined with the curve evolution theory,the application of image gradient vector flow,effectively overcome the lack of the snake model,make the research of active contour model is greatly extended.Study of active contour research problem in the field of medical image,in this paper,the level set model,geometric active contour evolution is studied.By combining the gradient vector flow and deal with the goal of multi-point spread curve geometric active contour is studied;A priori knowledge in the image information of active contour segmentation aspects of learning to do some research work;Based on clustering are studied with the geometric active contour segmentation method;Finally,study the figure optimization to minimize the energy function of level set model,and studied the application in the MR image segmentation.In this paper,the main research results are as follows:(1)An energy variational level set fitting algorithm named GVLS is proposed,which is based on the idea of global and local image structure information,and guide the level set model to capture the noise outside the small details,algorithm can complete the structure information of image target.PM filtering algorithm is introduced into the GVLS,the difference are used to calculate the image characteristics,and combine the characteristics of the local and global minimization algorithm model,eventually improve the ability of the level set model processing target image.(2)Gradient vector flow level set algorithm named GVFLS is proposed,the algorithm uses a new energy item to calculate the computational complexity and related binding minimization process of level set model.GVFLS based on regularization and curve curvature optimization thought,weigh the complexity of calculation,the segmentation accuracy is achieved by curvature estimation,level set model segmentation curve converges to the desired target.(3)An geometric active contour algorithm(GAC)based on priori knowledge and morphological characteristics is proposed,which is named MCS.In order to position target during the initial segmentation,introducing the midpoint circle Hough algorithm to determine circular structure of left ventricular image,and the priori knowledge based on CV model about the MCS is represented as velocity field,which is embedded into the iterative equation in the GAC.Under the guidance of a priori information model,MCS preliminary locates segmenting target boundary,and final driven segmentation curve converges to the target boundary.(4)An clustering segmentation based on geometric active contour algorithm is proposed,which is named KmGAC,and is simulates the partition curve of internal and external area,to minimizing energy function through iterative clustering algorithm,and learning foreground and background of image,make the curve evolution to minimize energy function values.(5)An multiphase level set segmentation algorithm based on graph optimization is proposed,which is named MLS.Improved level set method does not need to solve the Euler equation,also do not need to calculate any partial differential equation,and the minimizing of model uses the graph partitioning ideal.MLS has low requirements for target and parameter choice,and high stability,convergence rapidly.Finally,five kinds of algorithm on the same brain medical data are compared,and analyzed the algorithm of face data focus,analyses the segmentation effect and achieve the research purpose.