Segmentation Of Medical Images Based On Level Set

Segmentation of medical images plays a key role in medical imaging technology, and the effect of which will impact the results of computer assisted diagnosis. Recent years, a variety of algorithms has been proposed, most of which, however, concentrate on a specific issue.This thesis concentrates on the research on segmentation of medical images based on level set, which greatly promoted the development of active contour model. The cooperation of level set methods with curve evolution theory conquers multiple drawbacks of traditional snake model, and highly widens the application areas of active contours model.The thesis commences with comprehensive comparisons of algorithms available for segmentation of medical images. Then, geometric active contour model and level set methods are introduced, as well as two fast algorithms for level set methods before an evaluation of level set methods is made. Considering the large calculation amount of level set methods, it is presented in the thesis a variational level set algorithm for image segmentation, based on edge information, and without re-initialization, thus completely eliminating the need of costly re-initialization, accelerating contour evolution, and enhancing robustness. Moreover, considering poor evolution speed of level set methods, it is proposed in the thesis a level set segmentation algorithm based on sparse field methods, which reduces pixels to be renewed and increases time step each iteration, thus accelerating evolution, improving calculation accuracy, and applicable to such images with abundant data as medical images. The thesis concludes with the further research direction in future.