Active Contour Model Based Medical Image Segmentation And Applications

Traditional image segmentation techniques have many defects for complex image segmentation. Snakes, the active contour models, are widely implemented in medical image segmentation. The model is driven to the desired image features by an energy formulation, which consists of prior models and image data. GVF, Gradient Vector Flow, can deform the contour or surface into boundary concavities and expand the attracting area. The results of segmentation were bettered.This thesis presented further studies on GVF in three topics:(1) An improvement was performed on GVF to make the model more robust to the noise. The new method is called MSGVF based on scale space theory. A scale can be calculated by the scale of the last evolution. When the model is far from the interested boundary, a large scale is selected to increase the attracting area. On the contrary, a small scale is selected to make the convergence more accurate. Experiments proved MSGVF is more effective and more accurate than traditional GVF.(2) B-spline is used in boundary representation. Because continuity is implicit in the B-spline, there's no need to set several parameters of the parametric snake internal smoothness, such as weights for tension and rigidity. Another property of the B-spline is local control. This leads to an effective algorithm for computing the motion of the snake under the influence of image forces. And better results of segmentation were obtained.(3) Edge detection based on Canny operator is used to get the rough boundary of the interested area and get the 0-1 image. The control points are chosen from the rough boundary. It's easier to handle the procedures of initialization of contours.Summarize the contribution and innovation of the thesis as follow:(1) Propose a new Multiscale Gradient Vector Flow method of image segmentation. In this method, a new scale of edge map was calculated by the scale of the last evolution.(2) Use B-spline in the GVF method.(3) Propose an algorithm of image segmentation based on MSGVF and B-spline.(4) Propose a new active contour initialization method based on Canny edge detection to make the initialization easier and better.