Research On Geometrical Model Based Medical Image Segmentation

Image segmentation is the basic technology in image processing and computer vision, it is an important part of image analysis and visual systems. With the development of imaging technologies, the medical image is used to make a quantitative measurement and analysis before and after treatment, the analysis of medical image will help to disease diagnosis, follow-up of treatment or amendments to the patient’s treatment plan, which improve the accuracy and surgical success rate. Medical image in the medical application has a special significance. In this paper, we start to research the CT image, mainly using liver anatomical structures, we focus on the automatic segmentation method.First, the distance regularized level set method is researched in this paper. The content includes the distance regularization effect of double-well potential function, we demonstration DRLSE method in image segmentation application, which combines active contour model. The use of narrowband method to accelerate the evolution of the contour lines has been discussed. We research noise immunity of DRLSE methods of conducting experiments, which show the poor robustness of DRLSE to noise, we use nonlinear diffusion filter as supplement and improve DRLSE speed of evolution and segmentation accuracy.Secondly, we study the three-dimensional statistical shape model, describes the relevant statistical theory. Since that segmentation algorithm initial to locate the model manually will introduce errors, which limits its further application of three-dimensional. To simplify the positioning of the model, we propose an automatic positioning method by statistical histogram. In the description of the construction and application of the mean model, we presented a weighted average model as an innovation of theory.