High-Resolution Remote Visualization of Ray-Casted Volume Rendering on GPU Cluster
SESSION: Research Poster Reception
EVENT TYPE: Poster
TIME: 5:15PM - 7:00PM
AUTHOR(S):Jinghua Ge, Robert Kooima, Andrei Hutanu
ABSTRACT: This project studies high-resolution parallel volume rendering on GPU clusters. The parallel renderer is part of a distributed visualization system. The system features an interactive and collaborative remote visualization pipeline connected by high-speed networks. Other related features include progressive visualization and scalable video streaming. The resulting image is streamed to a local user’s desktop screen or large-scale display wall.
We bring out the concept of overlapping footprint (ofp) and implement the overlap-based image compositing schedule based on the SLIC algorithm. We also implement the image compositing operation on GPUs. These two practices largely reduce the image compositing cost of parallel volume rendering.
The poster will start with the system review and introduction of our remote visualization user community. Then our parallel volume rendering techniques are explained in detail. Performances analysis emphasizes on both weak and strong scalability study. We conclude with use cases, conclusion, and future work.