Image Segmentation Based On Level Set Method

Image segmentation is one of key issues in Computer Vision. Because of extracting only local information with disconnected boundary of the segmented region, and lack of ability to integrate prior knowledge about the segmented objects, classical non-model based image segmentation techniques cannot satisfy the requirements of complex image vision applications. In this case, a flexible framework is required that can integrate both low vision information from images and prior knowledge about target objects seamlessly to lead to a consistent representation of the segmented regions. Nowadays, the image segmentation based on level set method has received much appreciation, Such as the insensitivity to the initial curve position, the strong ability to deal with the topological changes etc.This paper have a study on image segmentation which is based on level set method. First , the methods, the target and the significance of image segmentation are introduced, and the Snake model, Mumford-Shah model, Chan-Vese model are discussed in detail. And then, the theory of curve evolution, how to solve the PDEs model based on level set method and its calculation methods are expatiated. One disadvantage of Level Set method is that the computational cost of Signed Distance Function is expensive. A new method—eight neighborhood Voronoi Source Sweeping, which can construct the Signed Distance Function fast,is present in this paper. This method possesses good accuracy and high speed. Compare our method with the direct method, the fast marching method and Lijun's method, the experimental result show that our method is efficient. Last, the Chan-Vese model can not get good edges of some multi-target images, so the energy term based on gradient is entered into Chan-Vese model to improve this model, the improved model not only make use of the image region-gray information but also make use of the image region-gradient information, and use the new method—eight neighborhood Voronoi Source Sweeping to construct the Signed DistanceFunction, so the improved model can get better result to the multi-target images, and also the using time of segmentation are shorting.