GPU Accelerated Voxel Traversal using the Prediction Buffer

The ever increasing size of data sets for scientific and medical visualization demands new isosurface volume rendering techniques to provide interactivity for the large datasets. The main obstacle to achieving interactivity is the computational bottleneck due to the dataset traversal and the corresponding amount of data transfer. We propose a novel GPU based dataset traversal technique that uses a prediction buffer to reduce the traversal time during dataset rotation. The reduction in the traversal time improves interactivity and consequently provides better insight into the dataset characteristics. We use a highly parallelized ray-casting algorithm and the proposed traversal technique to double the rendering speed. The factors which influence the rendering speed include block size, shared memory usage, and texture versus global memory. These factors were carefully considered to efficiently map the ray-casting volume rendering algorithm and the traversal technique to the GPU providing a high performance implementation. CR Categories: I.3.1 [Computing Methodologies]: Computer Graphics—Hardware ArchitectureI.3.7 [Computing Methodologies]: Computer Graphics—Three-Dimensional Graphics and Realism