Three-dimensional Medical Images Based On Mutual Information Registration And Snake Are Interested In Regional Integration

Since 1990s, the medical imaging technology has provided multi-modality medical images for clinical application, which can be classified into two classes, anatomy structure images(CT, MR) and function images(PET, SPECT). Medical image fusion can provide more information for clinical use, and benefit the clinical diagnoses, therapy and radiotherapy plan design, perfect the surgical operation and curative effect evaluation. Registration of medical image is the first step of image fusion, the precision determines the effect of fusion.The object of medical image registration is to seek a coordinate transformation, align the images in space position and anatomy structure. Fusion is to integrate the registered images into one image for more pathological information. Multi-modality head images are the research object of this paper, the research include: the pretreatment method, registration based on mutual information, segmentation of region of interest and fusion of ROL Finally, the registration precision by influence of rotation centroid is studied. Registration is widely used in digital image processing.Registration methods based on image gray have been proved to be the most robust and automatic method, the precision is higher than method based on feature and can reach sub-pixel precision. Registration based on mutual information is a typical one of these methods. It registers images by maximization of mutual information (MMI). The MMI method is present in this paper, the details are discussed all through this paper. The measure of mutual information is improved to reduce computation cost, the replace scheme of Powell's search direction is also rectified to reserve the original search direction, initial set of rotation can avoid local maximum induced by interpolation anti-facts, convergence threshold is set to reduce iteration region, morphology method is used to reduce noise of PET image, multi-resolution method is adopted to accelerate the registration. The results show that the modified MI measure can reach sub-voxel precision, and the registration time is reduced obviously.