Research Of Multi-modality Medical Image Registration And Fusion

Medical imaging technology has been widely applied in clinical diagnoses with the development of computer technology. There are many kinds of imaging modalities from the various imaging equipments. In clinic, it is necessary to integrate the different modalities into one image properly when the single-modality images can't provide enough information. The fusion methods can integrate the anatomical and functional information of medical images in different modalities into one image, and this image can show the tissues precisely and roundly for clinic applications. We need to solve the problem of image registration before image fusion. Image registration is the precondition of image fusion, and the result of image registration directly affects the qualities of fusion image.Based on the theories of mutual information, optimization and wavelet, this thesis makes a deeper research on multi-modality medical image registration and fusion. The main contents of this thesis are taken as follows:(1)The medical image registration based on mutual information is studied, including the mutual information principles, methods and steps. The causes of local maximum of the mutual information function are analyzed.(2)The local maximum of mutual information function is an objective existence. Powell, which is the most commonly used optimization algorithm, its performance is strongly dependent on the initial solution and disturbed by the local maximum. In this thesis, we proposed a hybrid algorithm combined by Powell and PSO, and a multi-resolution approach based wavelet is taken.(3)The medical image fusion based on wavelet is studied. According to the characteristics of anatomical and functional images, we selected suitable fusion rules for the Coefficient fusion at high frequency and low frequency separately. The results show that this method can effectively retain the useful information of the two images.(4)Proposed a pseudo-color medical image fusion algorithm based on IHS transform and wavelet transform.