A Implicit Similarity Method for Multi-Modal Image Registration

Image registration of different sensing types is of great interest in image processing area. Though lots of methods have been proposed, the results are still next to perfect. Of all these methods Mutual Information (MI) is a state-of-art one. However problems like incorporating spatial information still exists. Implicit similarity proposed by Keller overcomes some limitations of MI and obtained good results. In this paper, we proposed another implicit similarity method which resembled Keller´s. Our method tried to find the maximum posterior probability of a set of pixels in the source image on the reference image pixel set. Experiments with real optical and SAR images demonstrated our method´s feasibility.