J. Nonaka, Y. Watashiba, N. Kukimoto, N. Sakamoto, K. Koyamada, T. Iwashita, and M. Kanaza
Volume Rendering, Ray-Casting, Bounding Volume, Shared-Memory Multiprocessor.
This paper proposes a parallel extension of a Ray-Casting algorithm accelerated by an object-boundary projection technique. This technique is based on the bounding volume method that provides acceleration by skipping the empty space surrounding the objects within a volume data. By detecting and projecting the object's boundary as pre-processing, we can delimit the regions, both in image and in object-space, for effective ray firing and traversal. Focusing the implementation on shared-memory architecture parallel machines, the image-order partitioning scheme, wherein the screen is divided into blocks, was chosen. In order to improve load balancing, the object-space boundary information was used as a parameter for efficient workload distribution. We evaluated our software-based parallel rendering algorithm on a Fujitsu HPC2500, using 5123 medical volume dataset and 10242 , as the window size. The obtained results make clear the effectiveness of this approach because it reduces considerably the volume rendering time within some circumstances.
Important Links:
Go Back
