A novel memory access mechanism for arbitrary-view-projection volume rendering

Direct rendering of volume data is both conceptually simpler and more capable of producing visually superior results than other rendering methods. Special hardware architecture for direct volume rendering enables interactive exploration of volume data from arbitrary projection angles, which would otherwise take significantly longer due to its enormous computation and memory requirements. This paper proposes a new architecture called FAVOR, which features a novel memory access scheme that eliminates the most critical bottleneck in hardware rendering machines, and provides conflict-free parallel accesses to volume data. A preliminary performance analysis shows that the extra overhead incurred by this scheme is rather marginal, within 20% of the target performance for a 16 frames/s volume rendering rate at the 512 × 512 × 512 resolution.