The Image Guidance System Image Matching Algorithm

Some methods used in image registration are reviewed. Generally, some are based on image pixel gray value and some are based on image character. The algorithm based on pixel gray can realize the image matching accurately, but is strict to image quality. In the case of disturbance of light, the matching can not be so acute. The registration algorithm based on character of image is not sensitive to light and shelter, can overcome the non-line distortion caused by light difference, but the algorithm is complex.In this dissertation, after doing deep research on Sequential Similarity Detection Algorithm (SSDA), a two-step algorithm based on SSDA is proposed. Experiment demonstrates that the proposed algorithm has an improvement in the speed and precision compared to SSDA.As for algorithm based on image character, research on character selecting and distilling; similarity rule decision and realization is also done. Point character is selected as distilling unit, for shape point has enough information. some modification has been made to Sequential Similarity Detection Algorithm (SUSAN) algorithm. Image registration is realized according to Hausdorff Distance, and a fast matching program based on pyramid hierarchical searching algorithm is included.