Experiences on Parallel Visualization of Volume Datasets

This paper describes the work undertaken by our group on RT{based and projective{based parallel algorithms for direct volume visualization. It introduces the reader to direct visualization techniques and describes the parallel solutions we have designed to increase the interactivity of volume rendering. Image synthesis time is a critical issue for volume rendering techniques, because the practical every day use of visualization as an analytical tool to understand natural phenomena requires interactive dialogue between the user and the visualization system. We therefore focused our attention on the use of scalable parallel computers and of eecient parallel algorithms to reduce image synthesis times. Three diierent parallel algorithms are presented and discussed in the paper. The rst is a scalable ray tracing algorithm for rendering regular volume data. The second is a Delaunay triangula-tor which builds tetrahedral tessellations from unstructured datasets. The third is a projective{based algorithm to render thetraedra{tessellated volume datasets.