Research Of Image Segmentation Based On Level Set

Image segmentation is a classical and crucial problem in the fields of computer vision and image understanding. Traditional methods of image segmentation is difficult to meet the requirement of complex image segmentation. The image segmentation based on level set, is a kind of numerical method based on the theory of geometric active contour model, which is a simple, accurate and stable method, and had been applied extensively. However, there are some problems in the practical application of Level set image segmentation, such as the re-initialization, distance preserving and multi-target detection problem. The overall goal of this paper is to research these problems and propose solutions in order to make segmentation more accurate and effective.In this paper, we combine the theoretical study with the software development and analysis the problem related to the image segmentation of level set theoretically, and propose some new models, schemes and algorithms.At last,we have designed a software platform with property of either experimental or applied to verify its legitimacy by object-oriented programming. In the theory, we introduced the theory of active contour model,curve evolution and level set methods.we analyzed the problems of the region-based C_V model, the gradient-based Li model, and phase level set model, and researched the solutions. Based on the study of the typical level set model theory of the image segmentation,in the aspect of software development, we mainly use UML to analysis and design the software, use pseudo-code to describe the core algorithm, eventually the software has been implemented by Delphi,and a lot of testing has been done by the real medical images.Four results of innovative research and design are presented in this paper:①the edge pattern of automatic extraction of the initial contour based on the Li model had been proposed;②the mean progressive model of distance preserving based on the C_V model had been proposed;③the new scheme of multi-target detection based on Li model had been proposed;④the two patterns of the region adapt model had been proposed. The results of innovative research and design greatly improves the efficiency of image segmentation, and provides some powerfully technical assurance for the level set image segmentation can be applied in the variety of complex situations.