A Pipeline Algorithm for Interactive Volume Visualization

A parallel algorithm for volume visualization is presented which supports viewing of volume data from arbitrary directions. The parallelization is done by a multiple macro pipeline, because interactivity of the visualization process is the driving goal. An implementation on a distributed memory processor (nCUBE) with 64 nodes is shown, where the volume data set is cut into slices and distributed to the node processors. The processing power of this multiprocessor allows near interactive frame update rates for volumes of moderate size (128x128x128 voxels). Results of this implementation with special respect to speedup and scalability are discussed and further enhancements are proposed.