Research Of Medical Image Segmentation Algorithm Based On Active Contour Model

In recent years, the medical imaging technology has been in rapid development. The medical image processing is more and more important in modern medicine diagnosis, and the medical image segmentation is one of the key issues in computer-assisted medical diagnosis. The use of images to guide the surgery, radiation therapy for cancer etc., are all based on image segmentation. Since the medical image in essence has the characteristics of diversity and complexity, the results of the traditional image segmentation algorithm are less than ideal, making it easy to get the wrong segmentation results.The active contour model combines a variety of image information, and the contour extracted from the image is a closed curve. The purpose of the thesis is to find a medical image segmentation algorithm based on active contour model, which has a precise segmentation effect and can greatly reduce computational cost. The thesis firstly expounds the background of medical image segmentation, the significance of the research, and introduces the common medical image segmentation algorithm. Secondly, this thesis introduces the active contour model which has many advantages, and then put forward the key research direction:geometric active contour model. Thirdly, we divide geometric active contour model into two directions:contour-based geometric active contour model and region-based geometric active contour model, and in the end, we propose improved algorithms, respectively.Distance regularized level set evolution is a typical representative of the edge-based active contour model, which eliminates the need for re-initialization and thereby avoids its induced numerical errors. However, the model can’t very well overcome the contradiction between reducing the image’s noise and protecting the image’s edge imformation, and besides, adaptive segmentation can not be achieved. The thesis presents a new filter function, which can remove the noise while preserving edge information. Meanwhile, the introduction of the mechanism of adaptive evolution achieve the automatic evolution of the model. The segmentation accuracy and efficiency of improved algorithm are both advanced.C-V model is a typical representative of the region-based active contour model, which has good ability to handle the blurry boundary. However, it can not completely avoid re-initialization, and is also difficult to handle images with inhomogeneous intensity. The thesis adds a a penalty term to the C-V model, and proposes a new method to calculate the mean intensity value, and then introduces a mechanism to balance intensity unevenness inward and outward area’s curve. The experimental results indicate that the improved algorithm can handle the image with inhomogeneous intensity.