Research And Realization On Three-Dimension Visualization Of Brain MRI

3D Visualization of MRI in medicine is an important application of computergraphics and image processing in biomedicine engineer. It is widely used in thesefields such as diagnostic, surgery planning and simulating, and so on. Researching on3D visualization of MRI has significance in science and worthiness in practicalapplication.The main research contents of 3D direct volume rendering of MRI includingimage pre-processing, such as interpolating and filtering, 3D-classification andextracting tissues or organs of body, improving on direct volume rendering, and so on.In this dissertation, classification and direct volume rendering for 3D visualization ofMRI are studied, and a primary frame of 3D visualization system working on PC isdeveloped.In this dissertation, to ensure that 3D visualization can render the tissuesaccurately, two classification methods of 3D MRI are presented. The first one is aninteractive classification method of 3D MRI based on threshold of gray-levelinformation. It creates the threshold value and error range after analyzing thegray-level histograms of the whole 3D image and some slice images, interacts,binarizes the images, and processes the images using properly mathematicalmorphology operation according to the feature of tissues to extract, and so on. Thesecond one is a classification method of 3D MRI based on wavelet-domain HiddenMarkov Mode. First it introduces EM arithmetic to estimate parameter of HMT. Thenthe approximative class and the eigenvalue of each class in wavelet space are gainedby proper wavelet-decomposition. The result of classification is produced by ICMarithmetic.In this dissertation, to improve the quality of the visualization, two DirectVolume Rendering methods—Ray-casting and Shear-Warp are improved. (1) thehybrid method of central difference and adjacent difference to compute grads arepresented to solve the problem that the method of central difference results inerror when voxels changes acutely in Ray-casting method;A new composingmethod of sampling points is presented to solve the problem that blur occurs when theimage of 3D visualization is magnified partly. It integrates sampling,interpolatingwith compositing, saves the information of eight data points around sampling point,and converts longitudinal orderly operation of color composition into out-of-orderdistributing one in plane. (2) An improved Shear-Warp method is presented to solvethe problem that projective imaging method cannot show interface because it cannotgain the exact normal to compute the illumination effect. It introduces sphere andoblate sphere to express the distributing of all tissues in volume data after sheartranslation and to approach the normal of tissues interfaces by the normal of sphere,oblate sphere and the blending of them. So that illumination effect is computed andthe interface of the tissues is shown.