A New Automatic Band Registration Algorithm for Satellite Multispectral Remote Sensing Images

Abstract

Satellite multi-spectral images which captured from independent camera photography system of small satellite have nonlinear errors in spatial position.To this complicated problem,the paper introduces an automatic registration method to match multi-spectral images automatically,in which the mutual information,genetic algorithm and MQ geometric correction models are intergraded.In its algorithm,the mutual information is used as a similarity function of the registration problem,which can achieve an accurately registration precision;and the genetic algorithms are utilized to search the optimization value fast by its characteristics of finding global optima rapidly and efficiently;Meanwhile,the MQ geometric correction models are also adopted to develop a correspondence between the input and output images quickly.From the experimental results,it seems that this method can achieve a better accurate automatic registration results in bands nonlinear geometric relationships between multi-spectral images,whose error of the entire matching is less than a pixel.