Based On Geometric Active Contour Model For Image Segmentation Methods And Applied Research

Image segmentation is to separate an image into certain un-overlapped homogeneity regions which utilize certain characters of image, and a technique and process of extracting interesting object. It's a base of image analysis and image understanding. So the research of image segmentation methods has both the academic and application value.In recent years, active contour model implements image segmentation by mathematical model. The basic idea of active contour is:firstly, giving one curve or multi-curve; then defining an energy function about the image information and minimizing the function; finally, evolving the curve which will stop evolution at some characteristics of the image. Active contour models can be divided primarily into parametric active contour (PAC) and geometric active contour (GAC). The PAC model expresses the curve deformation with parameters form directly so as to unsolve the topological change. The GAC model based on curve evolution theory and level set method expresses the curve deformation indirectly with level set function which always keeps effective and continuous. So it allows topological change during curve evolution, and has a stable numerical solution. This grately develops the apply field of active contour model.We discuss the GAC model in this paper, and aim to do some research on image segmentation with GAC model to improve some existing models, and apply it to image segmentation on some gray and color images. Firstly, we introduce the basic theory of curve evolution and level set method. We also introduce some classical GAC models based on edge information. Secondly, we propose a multi-scale level set image segmentation combined of edge and region information based on analyzing C-V model and its advantages and disadvantages. Thirdly, we combine Bayesian theory based on maximum a posteriori and GAC model, and propose an adaptive level set algorithm for color image segmentation combined of region statistical information. The primary research contents and innovation of this paper include two sides which can be listed as follows:(1)We proposed a multi-scale level set algorithm for image segmentation combined of edge and region information. Construct an energy function combined of gradient and region information to get a hybrid C-V model, which constructs an edge detection function based on wavelet high-frequency components in gradient constraint term and applies region term of C-V model in region constraint term. Then solve it using variational level set method and eliminate the re-initialization procedure. The original image is first transformed into the wavelet domain to get a coarse approximation, and an approximation contour is obtained on the coarse approximation by the hybrid C-V model. The approximation contour is interpolated into the original-scale contour. Then the original-scale contour is taken as an initial level set function and the next active contour evolution which applies the C-V model of eliminating re-initialization is performed on the original image to get the real contour. Experimental results show that this method can extract object edge from image truly and efficiently, and has the advantages of anti-noise, un-sensitivity of initial contour, stable numerical solution. It is a perfect image segmentation method, and fit for the efficient segmentation of large images.(2) We proposed an adaptive level set algorithm for color image segmentation combined of region statistical information,In this model, we deduce image segmentation model with Maximum a posteriori according to Bayesian theory, and combine it into GAC model. Firstly, construct a speed stop function based on region statistical information which combines region statistical character and Bayesian model. And apply it to image segmentation on some color images. Secondly, construct alterable coefficient based on a posteriori of segmented image of inside and outside the curve. It aims to change evolution direction adaptively based on the information of image. Thirdly, eliminate the re-initialization procedure with introducing interior restrict energy term of the Li method. At the same time, simplify the definition of the initial level set function. Experimental results show that the proposed of speed stop function and alterable coefficient term promote the adaptability of evolving curve, and precision and efficiency of segmentation. When apply it into color image segmentation, this method has the advantages of evolving adaptively and stable numerical solution, and can extract object edge from color image truly and efficiently, so that is a perfect image segmentation method.