Deploying GPU-based Real-time DXT compression for Networked Visual Sharing
Namgon Lucas Kim, Jongwon Kim · Proceedings of the Asia-Pacific Advanced Network · 2011
The networked visual sharing application in multi-party collaboration environment needs compression of video streams due to network bandwidth limitation. For interactive real-time sharing, real-time compression of high-quality video as well as audio echo cancellation are required, which commonly depend on the availability of high-cost hard-to-setup specialized compression and echo-cancellation hardware. In this paper, by leveraging the computing power of GPU-accelerated PC (personal computer), we discuss how to support the software-only real-time compression of HD (high-definition) video streams. The chosen lightweight scheme, DXT (i.e., S3 Texture Compression), is highly matched with GPU-accelerated texture compression. By implementing GPU-accelerated DXT compression, based on CUDA (Compute Unified Device Architecture) parallel computing, and by deploying a software-based echo controller together, we can enable a low-cost solution for efficient networked visual sharing in collaboration environment.