Skip to Main Content
This paper presents a novel non-rigid medical image registration method to improve the quality in image registration based on feature. As DTCWT (Dual-Tree Complex Transform) is directionally selective, optimally decimated at coarse scales, invertible and has smoothly shift invariance, it is adopted instead of DWT (Discrete Wavelet Transform) in the algorithm presented. First multi-scale keypoints are detected using DTCWT, and then HD (Hausdorff distance) is used as the similarity measure between two point sets. Finally optimal affine transformation parameters are computed with NPSO (Niche Particle Swarm Optimization). The experiment results show the method presented has better robustness, antinoise and accuracy.