Registration of Satellite Imagery Using Genetic Algorithm
Rakesh Singhai, Jyoti Singhai · 2012
Image Registration is the process of determining a transform that provides the most accurate match between two images. The search for matching transformation can be automated with the use of suitable metric, but it is difficult to determine local maxima with direct search methods. In this paper a simple and powerful search strategy based on genetic algorithm is proposed to register satellite images. This method applies mutual information (MI) to measure statistical dependence of information redundancy between the image intensities of corresponding voxels in both floating image and reference image. Partial Volume distribution interpolation (PV) is used to compute the MI criterion. Scope of the paper is limited to pair of images, which are misaligned by rigid transformation (i.e. rotation and/or translation). Performance of genetic algorithm based proposed approach is compared with existing search methods, which shows that the proposed approach overcomes the limitation of local maxima with the desired accuracy and speed.