Design and Realization of Medical Image Nonrigid Matching Algorithm

Although rigid matching methods are a good robust point matching (RPM), they can not meet clinic needs. Because that many deformations are non-rigid and nonlinear, so we need non-rigid registration. But at present feature-based non-rigid registration methods for frequently encounter the correspondence and robust problem (H. Chui and A. Rangarajan, 2003). In this paper, based on thin plate spline and matching matrix, a matching algorithm for non-rigid medical image is presented, this algorithm decomposes the parameters of thin plate spline (TPS) into affine part and non-affine part, and solves the matching matrix through iterative means. Comparing with the popular iterated closest point (ICP) algorithm and other non-rigid matching algorithms, the proposed algorithm can not only keep biunique constraints between image feature points, but also is more robust