Font Size: a A A

Researches On Multi-Model Medical Image Registrtion Technique

Posted on:2012-04-08Degree:MasterType:Thesis
Country:ChinaCandidate:F ZhangFull Text:PDF
GTID:2218330362953069Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the development of medical, computer technology and biomedical engineering technology, medical imaging equipment offers a variety of models images for clinical diagnosis. Because of the different imaging principles and different imaging equipments, images express different information. Such as, anatomical images have the higher resolution, which provide anatomy information for organs; and functional images have lower resolution, which provides metabolic information for organ. If these two kinds of images are fused, the registered image would express richer information, which will greatly enhance the useful of image information in the clinical diagnosis and surgical therapy. The main work of this paper is as follows:Firstly, we introduce the background, significance and present situation of the medical image registration, described the problems faced and the development trends of the registration; then we deeply study the basic principles and the implementation steps of image registration.Secondly, we in-depth study the multi-modal medical image registration based on gradient of mutual information, then introduce the knowledge about mutual information and gradient information between images, according to the image registration method based on mutual information prone to the local optimization, we analyze the cause and propose the gradient of mutual information as similarity measure registration criteria, which can suppress local optimization effectively; the improved Powell algorithm is used to research the optimal parameters. Experimental results show that the method can suppress local maximum and search correct registration parameters easily.Finally, in multi-modal medical image registration method based on morphological Haar wavelet, we introduce the basic theory of morphology, and make full use of mathematical morphology which is the nonlinear filter essentially to preprocess the image, which can enhance the edge features and suppress noise. In this paper, we introduce the quantum particle swarm optimization and Powell algorithm to obtain the optimization parameters. Simulation results show that morphological Haar wavelet is used to process the images and hybrid optimization algorithm used to gain the optimization parameters, which can improve the stability and accuracy of the registration, and greatly reduce the computational cost.
Keywords/Search Tags:image registration, gradient of mutual information, quantum particle swarm optimization algorithm, Powell algorithm, morphological wavelet
PDF Full Text Request
Related items