Feature Points Matching of Nonrigid Tissues Based on SURF, Spatial Association Correspondence and Clustering: Application to MR 2-D Slice Deformation Measurement

Due to the nonlinear and nonuniform local deformation of the nonrigid tissues, it is difficult whereas important to extract and correctly match a considerable number of feature points from the MR images for deformation measurement. Current approaches are dissatisfying towards this issue. In this paper, firstly the authors use SURF algorithm to extract the feature points in the initial MR image, and take every point in the deformed MR image as the feature point. Then the SURF descriptors and Spatial Association Correspondence (SAC) of the neighborhood pixels is adopted to match the corresponding feature points between the initial and deformed MR images. Finally, by clustering the coordinate differences between the deformed points matched by SURF-SAC with the corresponding points calculated by affine transformation, most of wrong match points are eliminated. The experimental results prove that the proposed method can extract and match more correct corresponding feature point pairs than SURF and SIFT methods. Key words—SURF, Spatial Association Correspondence, Clustering, Feature point, Matching, Deformation