Based On The Characteristics Of Medical Images Of The Head Of The Contour Feature Extraction And Alignment

Fusion of Medical image is one of most important and useful subject in the field of image processing, while image registration is also the base of medical image processing. In this paper, after analyzing the imaging theory of medical image such as CT, MRI, PET etc, being acquainted with the level of image processing, a scheme is constituted to realize the registration between medical images, which is based on the feature point selected from the contour of head.In this research work, first the difference between the edge detection operators is compared, which is used on medical image of head. The Robert operator is selected after analyzing the result of edge detection. Threshold value is also been selected to change the gray image into black and white image, which can intensify the information of edges. Then thinning is used to enhancing the contour of image. Finally using contour trace arithmetic picks up the contour of object which is useful to the image registration. All the courses are all realized by software VC6.0. Furthermore, histogram, which can reflect the pixel value of image, and contour, which can be selected through inputting the position of the start point, are all realized by the software.Feature point abstraction and one to one correspondence of them is also been researched. A new way which based on the local feature point is created. As a result, the problem of imaging registration becomes more convenient. Single Value Decomposition, which is proved in the paper, is adopted to solve the registration. At last, the course of registration is emulated by MATLAB.In the support of above achievement, a flexible and exact way of medical image registration is found. The result in this paper can be used in many fields and high precision can also be gained.