Shared-memory multiprocessor implementation of voxelization for volume visualization

Shared-memory multiprocessor workstations have become widely available to the visualization community. Direct volume rendering of unstructured grids is a computationally intensive problem that is of substantial interest in scientific visualization. This paper presents a parallel algorithm for a voxelization-based direct volume rendering of unstructured grids and its implementation on a shared-memory multiprocessor. Two highlights of the algorithm are data coherence and work distribution. The data coherence property reduces voxelization to incremental computation of voxel values within a cell using a projection approach. The work distribution is achieved by multiple processors performing voxelization in parallel on a shared-memory machin