A Parallel Volume-Rendering System using Multiple-Shear Composition

Abstract

Fast direct volume-rendering systems are in high demand due to the increasing amount of scientific data generated by a variety of computer simulations, of medical data obtained by MRI and CT scanners, and of geological, oceanographic, and meteorological data collected from various sensors. A notable characteristic shared by these data volumes is the sheer amount of data to be rendered. A huge amount of computing power is thus needed for animated visualization, which is essential to observe many physical phenomena. We propose a fast parallel volume-rendering method using multiple-shear composition and describe its implementation on a PC-cluster. In this system, a volume is divided into subvolumes, one for each PC, and each PC generates base-plane image from its subvolume. These images are composited into a single base-plane image that is warped to get the correct image on the screen. Our proposed “multiple-shear composition” method is nearly four time faster than a raycasting method we implemented on the same PC-cluster. In addition, it has processing scalability due to the low volume of communications between PCs. A large volume can be handled because each PC handles only a subvolume. This system is thus suitable for large-scale scientific simulations.