This thesis aims at exploiting graphics processing units (GPUs) to accelerate real-time simulation and rendering methods. We target the GPU as the basic platform for several reasons. First, aside from their great 3D acceleration abilities, today’s GPUs offer tremendous computational power, high parallelism and flexible programmable ability. More and more researchers are turning to GPUs for better computation and rendering efficiency. Second, mapping the data and its computation kernels onto GPUs would greatly improve the rendering performance, since the updated data in GPU local memory can be directly used for visualization without extensive CPU-GPU data transfer. The main contributions of the thesis include: 1. A new method to simulate the hydraulic erosion phenomenon which runs at interactive rates on GPUs. The method is based on the velocity field of the running water. The velocity field is used to calculate the erosion process. The method has been carefully designed to be implemented totally on GPU, and thus takes full advantage of the parallelism of current graphics hardware. Results demonstrate that the proposed method is effective and efficient. It can create realistic erosion effects by rainfall and river flows, and produce fast simulation results for terrains with large sizes. 2. A fast and accurate level set framework on latest graphics hardware. Level set methods have been extensively used to track the dynamical interfaces between different materials for many scientific and engineering applications. Existing methods usually suffer from two problems: numerical diffusion and expensive computation cost. First, an improved level set method with better accuracy and important simplifications is proposed. Then each step is fully mapped on GPU with an innovative combination of different computation techniques. The proposed method successfully improves the computation accuracy, and provides real-time performance for large size 2D examples and moderate 3D examples. 3 A fast preprocessing method to remove the empty voxels from the volume data for better rendering performance. For most 3D volume data sets, a fraction of the volume is empty, which would bring down the rendering performance. A simple kd-tree based space partitioning scheme is proposed to efficiently remove the empty spaces from the volume data sets at the preprocessing stage. The splitting rule of the scheme is based on a simple yet effective cost function evaluated through a fast approximation of the bounding volume of the non-empty regions. The proposed scheme requires little preprocessing time and improves the rendering performance significantly. As a conclusion, this thesis provides some useful research work on real-time simulation and rendering methods on programmable graphics hardware.
修改评论