Normalized Mutual Information-based Image Registration Using Differential Search Algorithm

Image registration is the process of overlaying images of the same scene taken at different times, from different viewpoints, and/or by different sensors. Intensity-based image registration is considered to be a process of optimizing registration parameters of the similarity measure while different optimization methods are used to improve this process. Differential Search Algorithm (DSA) is a new nature-inspired optimization method with superiority at avoiding falling into local extremum. We studied the structure of DSA and applied it to the normalized mutual information-based image rigid registration. Experiments showed that compared with Particle Swarm Optimization (PSO), DSA always found the optimal solution and obtained stable results when registering different spatial transformations.