GPU-Based Volume Rendering Techniques

Visualization of three-dimensional data fields especially volume rendering has been the most important and rapidly developed techniques in the visualization of scientific computing in recent years. It has been developed based on some aspects of computer graphics, computer vision and computer image processing, and has been widely used in three-dimensional medical reconstruction, computational dynamic fluid, post-processing of finite element computation, seismic and geological applications.Volume rendering obtains final image by projecting original 3D data to 2D projective plane directly, without constructing geometrical graphic items, it has the advantage of displaying the different regions translucently and revealing interior details of data fields. As this technique requires dealing with huge amount of data, and has a comparably complicated algorithm for image generation, it is inherently time-consuming and costs greater computation than other algorithms such as surface rendering. Usually high-end graphic workstations and special-purpose hardware are used, it's difficult to be realized for reasonably sized data sets on general PCs at real-time and interactive frame rates. Fast evolution of PC hardware especially the programmable graphics hardware with its programmable vertex and fragment shader, have provided hardware acceleration support to realize real-time volume rendering.First, this paper gives an overview of the visualization of scientific computing and GPU. Then, analysis the main flow of visualization, do some pretreatment to 2D data slices to obtain correct volume data sets for our algorithm. In chapter 3, describe the key technique of volume rendering, compare some typical algorithms, realize the 3D texture volume rendering, and discuss its difficulties. 3D texture algorithm is equivalent to process all rays simultaneously in ray casting, to simplify calculation, adopt fixed view direction and frame cube, proxy geometry are rectangle slices which are perpendicular to the view direction and intersect with frame cube, the aim of change the view direction is achieved by rotating the texture coordinate, over operator is used to blend for the final image. In chapter 4, describe the graphic hardware pipeline, ultilizing the GPU's programmable vertex and fragment shader to change some render computations, like vertex transformation, texture mapping, color conversion of 3D texture volume rendering from CPU to GPU.Last, use VC++ as platform, OpenGL 1.5 as 3D API, and Nividia Coporation's high level shading language Cg to realize GPU-based volume rendering. The results show that, without compromising the image quality, the speed is improved.