Image Registration Based On Evolutionary Algorithm

Image registration tries to match two or more images which are about the same objective but taken at different situations, such as different time, different viewpoints, and/or different sensors. Image registration is one of the research hot topics in image processing. It also plays an important role in image fusion, change monitoring, environment monitoring and the classification of the multispectral image. Evolutionary algorithm is an efficient heuristic optimization algorithm. In this thesis, we convert image registration to an optimization problem, try to use evolutionary algorithms to obtain the optimal transformation matrix of registration. The major contributions are as follows:(1) We propose an evolutionary image registration algorithm based on pixel information or features. The chromosome and objective functions are defined for the registration problems. The experimental results on some natural images have shown that that simple target image registration is a feasible registration method with better performance in terms of the application and accuracy of algorithm.(2) We propose a multi-objective image registration method. The basic idea is to combine the pixel and features based methods and thus to take advantages of different registration methods. The experimental results have shown that it is a new and effective registration method.(3) We propose a pyramid image registration method for big images like remote sensing images. The basic idea is to reduce the running time of registration and increase efficiency of the algorithm. The experimental results have shown that while preserving the registration effect, the pyramid image registration method reduces the registration time, which means it is a registration method with higher time efficiency.