The research of image registration algorithm based on the combination of cross-correlation and matrix-multiply DFT

Technology of dual-band infrared image fusion can tremendously improve the reconnaissance capability for weapon equipment. While the detectors embedded in domestic dual-band thermography proves to be different, resulting in a lot of differences between the images obtained.When one wishes to fuse these images from disparate detectors, the first step is high-precision alignment. For the case of just a translation between two images, the usual technique to solve this problem is to compute the crosscorrelation by means of conventional FFT. In this paper the conventional FFT method is reviewed. Moreover, as compared to the method of this one, we give a method of combining cross-correlation and matrix-multiply DFT resampling, which consumes shorter time. At the same time, we give this method to dual-band infrared images registration, and through the platform of MATLAB simulation we achieve the accuracy of sub-pixel level.