Abstract: For the problem of registration which takes infrared image as the actual data and the visible light image as the referenced data,a multiresolution registration algorithm was presented based on non-subsampled contourlet transform(NSCT).The infrared image and the visible image are transformed into non-subsampled contourlet domain firstly,then the proposed method takes the gradient normalized mutual information as the analogous measure for the low frequency image of high scale and searches the registration parameters with the improved genetic algorithm(GA).The improved GA adopts the adaptive crossover operator and mutated operator and conquers the precocious phenomenon based on the description for the population grade of maturity.Therefore,the coarse registration result could be acquired.Subsequently,the low scale registration procedure is gone on furtherly under the guide of the coarse result and the whole registration result would be obtained in the end.The experimental results show that the proposed method has high registration accuracy and fast speed.

Key words: Image registration,Infrared and visible light,Non-subsampled contourlet transform,Gradient normalized mutual information,Genetic algorithm