Based On Mutual Information Of Multi-modality Image Registration

Image registration based on mutual information (MI) has become an increasing popular match metric due to its wide applicability and overall accuracy. In this paper, image registration based on mutual information is discussed mainly.This thesis firstly introduces the major registration algorithms. Then the process of registration based on mutual information (MI) is described.How the interpolation process may affect the mutual information between images is discussed. Local maxima in multimodality image registration based on mutual information have a large influence on optimization. Based on the partial volume interpolation, the method of kernel function is introduced to decrease the local maxima. Simulations have been done to illustrate that local maxima are eliminated to a great extent.To further accelerate registration speed and accuracy, we design and realize wavelets based coarse to fine image registration method.The problem of manually choosing point for non-rigid image registration is discussed. We make use of template matching to automatically choosing point. Thin plate interpolation is used to realize the global elastic registration. Simulations have been done to illustrate that the efficiency and accuracy of this method in registration strategy.