Two Kinds Of Image Segmentation Based On Level Set Method

Image processing is the basis of computer vision and the important component ofimage understanding as well. With the development of computer, image processing hasbeen paid more attentions and has developed rapidly recently. It has penetrated into allaspects of living and social development.Image segmentation is a classical and crucial problem in the fields of computer visionand image understanding, it takes a crucial position in image processing. With more widelyapplications of image segmentation to every field, traditional linear methods have displayedsome limitations. Recently, the nonlinear methods, especially image segmentation methodsbased on PDE have been paid more and more attentions. Among PDE-based imagesegmentation methods, level set method is a useful and important one, based on whichactive contour models embody the advantages of PDE-based image segmentation methods. However, it has disadvantages, for instance, when simulating images with the traditionallevel set method, one use the average of gray values of the region as coe?cients, but thisapproach is not so e?ective when processing image with asymptotic gray average.In this dissertation, we mainly discuss image segmentation based on the level setmethod. After reviewing the development and current situation of digital image processingand image segmentation, we introduce the theory of active contour models, curve evolutionand level set method.Then we propose an improved level set method, which combine the level set methodand the watershed segmentation method in mathematical morphology, modify the coef-ficients of level set and make the model to segment the image with slight characteristicsand complexity. This method can take advantage of the globe information of the imageand obtain satisfactory results in image segmentation.In the end, we present a image segmentation model combining the level set methodand region competition segmentation model. Results of the segmentation experiments prove that our model is feasible.