Hardware-accelerated volume rendering based on DXT compressed datasets

Currently, large-scale scientific and engineering simulation has reached terabyte scale, and the resulting time-varying volume data present great challenges to GPU-based volume rendering. Because huge data can not be fully loaded into the GPU memory, real-time data exchange between disk and memory is inevitable, which greatly reduces volume rendering frame rates. This paper presents a novel GPU-accelerated volume rendering algorithm, which makes use of the DXT (Data Extract) compression technique in both data preprocessing and the volume rendering. Through spatial compression strategy, the large-scale volume data at a single time step can be rendered at interactive speed. When applying the data compression in temporal domain and hiding the disk access latency during rendering, the rendering frame rates of medium-scale time-varying datasets can be significantly improved. Experimental results show that the image difference between the original data and the compressed one can be well adapted to the visual acceptance of researchers.