Research Of SVC In Medical Image Registration Application

Medical Image Registration is an interdisciplinary area of information science, image technology of computer and contemporary medicine. It has already been used in many applications such as clinical treatment and preoperative diagnosis. However, the result of registration is not ideal because of the impact of noise, distortion of image detail and ambiguity. To solve the problem SVC is introduced in the paper. As a clustering method of attention, SVC has high accuracy and speed, which is a good solution to the question above. The major works are as follows:(1) This dissertation first studied the image registration based on feature. Both feature points or contour in the image registration process is just focus on the part of the image information. Registration based on contour can realize registration quickly, but is less able to deal with the regions which are away from the contour. Registration based on feature points has highly flexibility and it can be distributed in any location of the image, but is less able to control the overall deformation.(2) Image preprocessing has a great impact on registration accuracy and robustness. Noise signal makes some borders of organizations become blurred and fine structures difficult to distinguish. The paper presented a medical image denoising method based on SVC with multi-window. Experiments show that this method can not only remove noise, but also keep the details and low ambiguity well.(3) A Medical Image Registration based on SVC was proposed guiding together with contour and point feature. First, SVC was used in the contour point set which is extracted crudely to determine the key points and keep the image borders clearly; Second, SVC was used in the point feature set to make sure the location of the point and improving the computational efficiency; Meanwhile, the cumulative sum of distance between forward and positive directions has been used as similarity measure, so that to improve registration accuracy and robustness and to compensate the differences between the images.