Estimation of volume rendering efficiency with GPU in a parallel distributed environment

Abstract

Visualization methods of medical imagery based on volumetric data constitute a fundamental tool for medical diagnosis, training and pre-surgical planning. Often, large volume sizes and/or the complexity of the required computations present serious obstacles for reaching higher levels of realism and real-time performance. Performance and efficiency are two critical aspects in traditional algorithms based on complex lighting models. To overcome these problems, a volume rendering algorithm, PD-Render intra for individual networked nodes in a parallel distributed architecture with a single GPU per node is presented in this paper. The implemented algorithm is able to achieve photorealistic rendering as well as a high signal-tonoise ratio at interactive frame rates. Experiments show excellent results in terms of efficiency and performance for rendering medical volumes in real time.