Cvh Image Segmentation Based On Active Contour Models

Medical image segmentation is the key of realizing the VHP. Now active contour models are the focus of image process and computer vision. The primary task of the thesis is to study image segmentation's methods based on active contour models which are suitable for the anatomical images of the first Chinese female visible human(CVH),and to segment the bone. The innovative achievements of the thesis include:1. A new segmentation method based on GVF snake is proposed. By computing GVF on different scales and passing the final contour of the coarse scale to the fine scale as the first one, it can reduce the total times of evolvement and 60% of the algorithm's compute time than the original segmentation algorithm based on GVF snake. And a method of using active contour model on color images is proposed, which can utilize the color information of the image more adequately than the method using the GVF snake model on the gray image directly.2. A improvement segmentation method based on Mumford-Shah model is proposed. Using anisotropic diffusion before evolvement, it can solve the problem that simple Mumford-Shah model cannot process the complicated images including yawp perfectly without complicating the model. The method successfully introduces the multi-scale and multi-initialization strategies to compute the energy of the Mumford-Shah model, and reduces 70% of the algorithm's compute time than the segmentation algorithm based on C-V simple Mumford-Shah model.