Research On Selective Image Segmentation Based On Active Contour Model

As an important part of computer vision and image processing,image segmentation has been a difficult and hot research topic in the field of image engineering in recent years.The main purpose of segmentation is to divide the whole image into several non-overlapping subregions,and pixels in each region share a certain commonality.Because of its excellent extensibility and relatively complete mathematical theory,the theory of partial differential equation has become one of the most popular theories in the field of image engineering quickly.The active contour model based on level set method is a typical application of partial differential equation theory in the field of image segmentation.The basic idea can be summarized as following.Firstly,the evolution curve is represented by a highdimensional embedding function,and introduce it into the energy functional,and the functional of the embedding function is obtained.Secondly,a partial differential equation of the embedding function can be obtained by variational method,and the final evolution result is the boundary of the target image.After analyzing and studying the advantages and disadvantages of the existing models,the selective segmentation models and the dual level set segmentation model for intensity inhomogeneity,weak-edge and low-contrast images are proposed in this paper.The main contributions of this dissertation are given as follows:1.For the selective segmentation model based on edge information can not segment intensity inhomogeneity,weak-edge and low-contrast images effectively.The local region information of the image is added as a constraint,and a selective segmentation model for intensity inhomogeneity images is proposed in this paper.This model utilizes the local fitting values to drive curve evolution,to implement the selective segmentation of intensity inhomogeneity,weak-edge and low-contrast images.Simulation results show that the improved model can selectively segment images with intensity inhomogeneity,weak-edges and lowcontrast.2.In order to solve the problem that selective segmentation model for intensity inhomogeneity images is sensitive to the initial location of the evolution curve and the segmentation time is long,the further improvement is made in this paper.Firstly,a weighting function about local region information is constructed.Secondly,the global information of the image is introduced into the model.The weighting function controls the proportion of local and global region information in the model automatically according to the local fitting value of the image,and achieves the adaptive and selective segmentation of images ultimately.This improved model can make up the shortcomings of each other by using the advantages of the global and local region fitting,and finally realize the adaptive and selective segmentation of intensity inhomogeneity,weak-edge and low-contrast images.Simulation results show that the improved model no longer sensitive to the position of the initial contour.Meanwhile,the speed of segmentation is improved and the weight coefficient is no longer need to select manually according to experience.3.Since the improved active contour models are based on the single level set method and the local selective segmentation can be achieved only.In order to realize global segmentation and local segmentation simultaneously,this paper gets a dual level set segmentation model by defining two evolution curves.One curve evolves to locally segment image and the other curve evolves to globally segment the image.