Multimodality Medical Image Registration Based On Radon Transform

Image registration is the basic of image fusion, it is widely used in computer vision, remote sensing, movement estimation, medical analysis and many other areas. Medical image registration and fusion is an indispensable part of modern medical treatment. Medical image from different imaging equipment has its special characteristics, the registration of different modality medical images could integrate the anatomic and functional information of the pathologic structures and organs.The medical image registration algorithm proposed in this paper is mainly divided into two parts, one is coarse registration and the other is precise registration. The centroid of image is used to estimate the translation parameter, and the Radon transform is used to estimate the rotation parameter. The translation and the rotation are optimized as the initial parameters. The metric of the precise registration is combining normalized mutual information and gradient similarity. Powell is used as the optimization to calculate the final parameters.This paper introduces the significance, application and development of medical image registration first. Then it summarizes medical image registration briefly, introduces the basic methods of registration and analyses every part of the registration process, for example, the space transform, the interpolation, the metric and the optimization and so on. Then it introduces the concept of entropy, mutual information and the principle of registration algorithm based on mutual information. At last it introduces the innovation of this paper and introduces the concept of Radon transform and how to use Radon transform to estimate the rotation parameter, then it introduces the definition of gradient similarity and the metric used in this paper which combines mutual information and gradient similarity. The result of the experiment shows the validity and the veracity.