GPU-Based Adaptive Processing And Visualization Of Tetrahedral Volumetric Datasets

Volume rendering is one of the most popular visualization techniques in the field of scientific visualization. Roughly speaking, the datasets can be categorized into two classes: rectilinear volumetric datasets, and irregular volumetric ones. Among the latter, the tetra-hedral volumetric datasets have played important roles in many applications such as CFD, cosmology and etc. But their visualization remains a challenge. With the development of GPU technology in recent years, utilizing GPU for acceleration is one of the hot topics. To take full advantage of the parallel computing ability of GPU, we propose a set of tech-niques to achieve adaptive processing and visualization of tetrahedral volumetric datasets. In particular, we propose two regularization approaches for both sparse volume datasets and non-sparse ones. For the former, a dual-structure construction is employed to decompose the dataset into two components, which can be represented as two GPU-friendly textures. When the dataset is not spare, an adaptive mesh refinement process is performed to char-acterize the density distribution in an adaptive fashion. In each way, the new data structure is memory-efficient, GPU-friendly, and thus allows for effective GPU-based volume visu-alization. We verify the feasibility and usability of our method with various experiments.