Research Of Level Set Image Segmentation Method Based On Prior Knowledge Constraint

Image segmentation, which is a basic and important part of image processing, is the premise of image understanding and target recognition. The segmentation results directly affect the quality of subsequent image processing. Because the importance of image segmentation, researchers have proposed many segmentation methods and techniques, and there are lots of new research results every year. But existing methods still have some blemish, such as Complex segmentation process, low accuracy and poor generality.In recent years, Level set methods become a hotspot in the image segmentation. Level set methods have the advantages of free topology changes, strong ability of curve approximation and high segmentation accuracy. Although the level set method show a good segmentation performance, but the level set method is proposed as a kind of general image segmentation model, cannot applicable to all types of images, and the segmentation accuracy affected by initial and the set of parameters.This paper mainly takes the level set model based on local statistical information and the prior information of image. The main works can be summarized as follows:(1) On the basis of LCV(Local Chan Vese)model, import the shape and color information as prior information guide the evolution of segmentation curve. Shape prior information can guide the segmentation curve evolution to the target contour and aid the segmentation problems involving missing data or occlusions. Color consistency constraint used to determine the similarity between target and prior image, improve the accuracy of segmentation. Specific to the complex background images, which have fuzzy boundaries, covered and the ones having the similar gray value with background, have rapid and accurate segmentation.(2) Generally, shape based level set method must extract the prior shape before segmentation, based on the prior shape to segment the similar target, generality is poor. Shape symmetry is a useful shape feature of object, which widely exist in nature and artificial objects. Use the shape characteristic of symmetry as segment constraint conditions to segment the target with shape of symmetry in the image. So this paper treats the symmetry as a shape constraint conditions, the main emphasis is to design a constraint function to detect symmetry, and then integrated this constraint function into level set method, then build a new and effective symmetry based level set methods for image segmentation.