Research On Key Technologies For Graphics Hardware Assisted Volume Rendering

Volume rendering for the 3D volumetric data is the most important and rapidly progressing technology of scientific visualization.Volume rendering which combines the knowledge of Computer Graphics,Computer Vision and Digital Image Processing has been widely applied in fields of Medical 3D Reconstruction,Computational Fluid Dynamics,Finite Element Analysis and Geological Exploring.Intermediate geometry primitives are not necessary in volume rendering.During the process,volume rendering extracts meaningful information directly from volumetric data then maps into a single 2D image.Volume rendering convey more information than surface rendering images,but at the cost of increased algorithm complexity,and consequently increased rendering times.Not long ago,real time interactive volume rendering is beyond the computational capacity of PC hardware,and usually only can be realized with the support from the high price graphics workstations or specialized graphics hardware.Recent years,with the boost of processing power and programmability in the consumer graphics,it's possible to implement real time volume rendering system on the PC platform.We put much effort to develop the hardware accelerated volume rendering method to utilize the high performance parallel computation model of current Graphics Processing Unit(GPU).This thesis has three parts:Part 1.We introduce the theory basis of classic volume rendering.The important formulas of volume rendering were carefully studied.Then we compare the traits of several commonly used volume rendering methods,and decide to develop new acceleration method for ray-casting.Part 2.The architecture and programming model of GPU are described.The method of porting code from CPU to GPU is also introduced.Part 3.We address the integration of texture based volume rendering on GPU.Then, we describe single pass ray-casting on programmable graphics hardware as an alternative to object-order approaches.We exploit the octree based hierarchical acceleration structure to skip empty space along the rays of sight.We also do some research on the optimization of shading language code and using 32bit precision depth buffer to improve rendering quality.We demonstrate performance gains up to a factor of 4 for typical renditions of volumetric data on the NVIDIA 7900GS graphics card. We conclude that our volume rendering acceleration method can be used to achieve high quality real time volume rendering with standard PC hardware.