Fast image registration algorithm for compound IR/visible light homing guidance

To overcome the limitations of large computational cost in current multi-sensor image registration algorithms for compound IR/visible homing guidance, the authors proposed a fast method based on partial Hausdorff distance and a definition of similarity measurement for two groups. This method has realized a fast IR/visible image registration algorithm that effectively decreases mis-matching of feature point pairs. First, this algorithm extracts remarkable feature points from the image edges; second, based on the proposed fast method and the similarity measurement definition, it roughly estimates rotational and translation parameters; thirdly, according to the estimated parameters, it narrows the search scope in matching feature points and choosing the matching feature pairs; finally, it calculates the optimum transformation parameters by the least square method. Simulation results indicate that this registration algorithm is not only effective, but also fast.