The Research Of Image Segmentation Method Based On Mumford-Shah Model

As the information technology gradually blends into human life, the image information becomes the important information that human get, and how to use the image information becomes more and more important. Image segmentation is a important and foundational problem in the field of image processing, and has a large academic and practical significance. Although there are a lot of methods to image segmentation, some problems are still need to be solved.Level set methods based on Mumford-Shah model are the excellent and important methods based on deformable model. Because of depending on global information of homogeneous regions in the image, they segment the image quickly and precisely. But the experiments show that traditional level set methods can not segment the human brain MRI (Magnetic Resonance Imaging) images.And the traditional level set methods must re-initialize level set functions costly so that level set functions can be close to a SDF (Signed Distance Function) and image can be effectively segmented. But in order to be close to a SDF, the time step must be small, and the evolution procedure is slowed down, especially in the large images.This thesis improves traditional level set methods based on Mumford-Shah model by introducing the method of level set without re-initialization into them. Besides remaining the global optimization which traditional level set methods own, the modified method fastens the evolution procedure, and can segment human brain MRI images or similar images, while the traditional level set methods can not. The experiments indicate that typically all edges of an image will be picked up only no more than 10 iterations by using the proposed method. The segmentation tests for synthesized, biological and medical images prove that the proposed method is very quick and robust.