Segmentation Of MR Image Based On Information Entropy

With the development of medical imaging technology, biology medical image segmentation technology has very important application in the diagnosis and treatment of patients in clinical medicine. However, because of medical imaging equipment limitations, medical image itself organizational differences and the impact of human factors, it lead to the images which we obtained include of noise, weak boundary and bias fields etc. This makes the traditional medical image segmentation method cannot obtain ideal segmentation results. At present, Active Contour Models (ACM) is a hot research direction of image segmentation technology for the rigorous mathematical theory supporting which based on level set method.There are two ordinary segmentation models, which is edge-based ACM and region-based ACM. The edge-based models usually utilize the gradient information of images to segment the specific target, but in generally images contain noise, weak boundary or boundary missing, which lead to segmentation results are easy fall into local optimum or over segmentation. In contrast, the region-based models are used widely. But due to the bias fields effects of images, which lead to segmentation results are not ideal.According to the research, the gray value of most medical images exhibit Gauss distribution, so we can utilize the methods based on statistical information. In this paper introduced the information entropy that can better to overcome the image noise, weak boundary and bias fields. The main research contents of this paper including the following two aspects:(1) This paper proposed a GAC model based on space information entropy. This method utilize the gray statistic information based on global image, which can well overcome noise, bias fields, weak boundary and can better to avoid the local optimum. This method also can adaptive expand or shrink to the target boundary.(2) This paper proposed a GAC model based on local information entropy. This method utilized the entropy based on statistical information in local domain can well describe the distribution of gray values. At the same time, this method utilizes symbol pressure function can adaptive drive the closed contours converge to the object boundary.