An Efficient Parallel Volume Rendering Algorithm Based on Shear-Warp Factorization on Distributed Memory Multiprocessor Systems

Abstract

Volume Rendering is a technology to illustrate the shape and volumetric property of the objects in the areas like medical imaging and computational fluid dynamics. Because volume rendering takes a lot of time to process a large volume of data, it is very difficult to generate three-dimensional image in real-time. In this paper, we present an efficient parallel volume rendering algorithm based on shear-warp factorization for medical imaging on distributed memory multiprocessor systems. We first use a novel data partition method to distribute voxels into each processor element and to render partial subvolume images independedtly. In the reassembling stage,it only requires merge time to produce the final image. Our parallel volume rendering algorithm takes less memory space and rendering time on distibuted memory multiprocessor systems. In our implementation, it can render a medical datasetof 256x256x256 voxels at about 75 mini-second per frame on a 32-processor IBM SP2 distributed memory multiprocessor system. Therefore, our algorithm can be used to make real-time volume rendering.