Slab-Based Raycasting: Exploiting GPU Computing for Volume Visualization

The ray traversal in GPU-based volume raycasting is usually implemented in a fragment shader, utilizing the hardware in a way that was not originally intended. New programming interfaces for GPU computing, such as CUDA or OpenCL, support a more general programming model and the use of additional device features, which are not accessible through traditional shader programming. In this paper we first compare fragment shader implementations of basic raycasting to implementations directly translated to CUDA kernels. Then we propose a new slab-based raycasting technique that is modeled specifically to use the additional device features to accelerate volume rendering. We conclude that new GPU computing approaches can only gain a small performance advantage when directly porting the basic raycasting algorithm. However, they can be beneficial through novel acceleration methods that use the hardware features not available to shader implementations.