Research On Hierarchical Image Registration Algorithm By Multiple Search Strategies

As the application area of image processing enlarges, information provided by single sensor can not satisfy the practical need. In this case, multi-source image fusion technology has been developed rapidly. Due to the different optical path and principle of different sensors, the image may have linear or nonlinear distortion with another as it is acquired at different time, from different angles and in different environment.So it is necessary to register these images before image fusion.The registration algorithms may have high automation and robustness, good adaptability, quick speed and high accuracy. Accuracy, speed and robustness are the most important specialties of image registration.However accuracy and speed are incompatible. Thus, in this thesis, we try to find the efficient image registration algorithms which give consideration to these two characteristics, and also satisfy the need of robustness.Our work in this thesis is mainly about:1) After analyzing the research in the domestic and abroad field of image registration technology, two common search strategies which are search stratagy based on Fourier-Mellin transform and search stratagy based on mutual information are compared.The experimental results show the advantages and disadvantages of these two search strategies;2) From the characteristics of search stratagie based on Fourier-Mellin transform and based on mutual information, algorithms based on hierarchical technology which give consideration to regiatration accuracy and regiatration speed are discussed.We design two hierarchical registration algorithms using multiple search strategies which are pyramid algorithm based on Fourier-Mellin and mutual information, and pyramid algorithm based on mutual information and Fourier transform. Then, the experiments based on single-modality images and multimodality images are simulated and compared.3) The robustness of these two algorithms is simulated based on single-modality and multimodality images which are added with some kinds of noises.