Key Techniques Study On Interactive Volume Rendering And Applications

Volume visualization especially volume rendering has been the most important and rapid developed techniques in the visualization of three-dimensional data sets in recent years. However, there are still some difficult issues need to be studied further, including several key techniques: new accelerating methods, new rendering improvement techniques, flexible setup and tuning methods for rendering parameters, which are the main topics of this paper. First, because traditional ray casting algorithm suffers from high computationalcomplexity and low rendering speed, a new ray casting accelerating algorithm is presented in this paper. The main idea of the new algorithm is to reduce the computing time of the resampling process by adopting a pre-gradient estimation, pre-classification and coloring, pre-shading schem. The new algorithm can efficiently accelerate ray casting algorithm without trade-off image quality obviously. Second, according to the O ( N 3) computational complexity of Splatting algorithm, a run length encoding (RLE) accelerated, pre-classification and pre-shading Sheet-Buffer Splatting algorithm is presented in this paper. This new technique saves rendering time by employing RLE mechanism to skip over transparent voxels. A fast and accurate Sheet-Buffer Splatting method is used in the rendering process, which accelerates the Splatting by traversing both the voxel scanline and the image scanline in Sheet-Buffer simultaneously. This is the first time in the research society to apply RLE techniques into the acceleration of object-order volume rendering algorithms. Third, we find that Splatting suffers from several kinds of aliasings because of fundamental principle inaccuracy. As an improvement, a new View-Buffer based Splatting algorithm is demonstrated in this paper. The new algorithm is based on View-Buffer rather than Sheet-Buffer. By using a voxel traversal order which is parallel to the view-buffer, and a pre-integrated three dimensional lookup table of reconstruction kernel, the method can not only solve the overlay and the popping problem, but also reduce the side-effect of integration problem to the rendering image. Fourth,in order to optimize a key process of volume rendering: the resampling process, a new object-order volume rendering algorithm is proposed in this paper. The new algorithm inherits all advantages of the object-order methods. Because the new method can make full use of the speedup bringing from pre-processing, at the same time the resampling operation in the proposed method is much coincidental to the sample theory, it can obtain higher rendering quality than Splatting with only a very little time cost. Local data features play important roles in the transfer function design for volume rendering. By examining relationships among three eigenvalues of inertia matrix, a semi-automatic data-driven transfer function design method is presented in this paper. This is also the first time to apply local shape features in the volume data into the guideness of the transfer function design. Furthermore, the proposed method can objectively fulfill the function of the previous boundary based method and extend the domain of transfer function to include more data features. At last, a three-dimensional medical reconstruction system---VolGraph is developed. Besides traditional medical image processing functions, the system also integrates several volume visualization algorithms including Marching Cubes, Ray Casting, and Splatting.