Vizard II, a PCI-card for real-time volume rendering

In this paper we present a second generation VIZARD system being capable of rendering 256 3 datasets at interactive frame-rates providing high image quality. In contrast to the previous VIZARD system, we use dedicated memory to store the dataset on the PCI card. Lossy and redundant compression of the dataset has been eliminated. Interactive c hange of shading and classiication parameters is enabled by m o v-ing shading and classiication from pre-processing into the pipeline. Memory bandwidth requirements are reduced by using a table of pre-calculated gradients. Thus, the gradient at sample location requires an eight neighborhood of voxels instead of a 32 neighborhood. We describe the generation of the discrete gradients and the impact on image quality. Furthermore, we present a parameterization of the ray in order to remove work-load from the pipeline. Finally, w e propose a PCI card serving as a platform for our second generation VIZARD. The proposed PCI card uses programmable devices to enable the implementation of other hardware accelerators as well.