Vector quantization for volume rendering

Abstract

Volume rendering techniques typically process volumetric data in raw, uncompressed form. As algorithmic and architectural advances improve rendering speeds, however, larger data sets will be evaluated requiring consideration of data storage and transmission issues. In this paper, we analyze the data compression requirements for volume rendering applications and present a solution based on vector quantization. The proposed system compresses volumetric data and then renders images directly from the new data format. Tests on a fluid flow data set demonstrate that good image quality may be achieved at a compression ratio of 17:1 with only a 5 percent cost in additional rendering time.