Research Of Nonlinear Medical Image Registration Based On Edge-Detection

Image segmentation is one of the classical and difficult question for research in the area which include image process, image analysis and computer vision, the medical image segmentation is always the challenge question in the computer vision area for the variety of the shapes of medical images.The Image Registration is a new researching field developing in the middle 90's of 20 century, also it is the important aspect in the research and apply of computer graphics and image analysis technology area. It is easy for people to study our bodies better, and can locate organs and focus accurately. It has good foreground and can apply widely.In traditional active contour method, an initial curve which called an active contour is placed somewhere near cortex. By modeling this initial curve as a physical object, an iterative procedure is initiated to cause the active contour to move toward the real edge. Active contour algorithm is traditionally developed by formulation and function that include two parts of internal energy an external energy, and should minimize the energy function of the formulation. The traditional method offers several advantages. However, in medical images that have complex structures, the evolution of the active contour will be seriously affected. In this paper, we present a new active contour model which include two new energy terms. The external energy term that we proposed can make the active contour toward the edge of the cortex properly. The new internal term can reduce time that used in computation.The medical images registration and fusion are very effective in image process. We can't only do this in different cross-sections of same patient, but also do it between different patients, and images of patient to standard atlas. In medical images registration, there are two ways that called rigid and nonlinear transformation to get to destination. It is suit for rigid transformation for same part of same patient. There are more application in another situation. In this paper we use elastic registration to process two images between different cross-sections of the same patient. After we get contour of the two images using active contour method, we propose a new algorithm for getting one-to-one mapping points. Then we get good result using EDT algorithm for two matching images. All the work provides the good foundation for fusion of multi-modality and 3-D reconstruction.