一种基于sort-last架构的非结构网格并行体可视化算法

A sort-last architecture based parallel volume visualization algorithm for unstructured grid

体绘制是非结构网格数据体可视化的重要途径之一。然而,现有的串行非结构网格体绘制算法效率较低,难以满足大规模数据体可视化需求。因此,并行体绘制算法研究是高性能体可视化程序设计的趋势。提出一种基于sort-last架构的非结构网格并行体绘制算法。首先,设计了基于KD树的并行体数据分割算法;其次,每个进程使用独立的可视化管线计算体绘制图像,根据树合成策略合成最终结果;最后,采用两层LOD模型技术优化交互体验。实验结果表明,提出的算法能被很好地应用于大规模非结构网格数据体可视化中,所有交互时延在毫秒级别,满足了实时交互需求。

Volume rendering is one of significant unstructured grid data volume visualization methods.However,existing serial algorithms for unstructured grid volume rendering are inefficient and cannot meet large-scale data volume visualization needs.Therefore,parallel volume rendering algorithm research has become a trend in high-performance volume visualization program design.A sort-last based parallel volume rendering algorithm for unstructured gird was proposed.Firstly,a parallel KD-tree algorithm for volumetric data split was designed.Secondly,each process calculated volume rendering images with independent visualization pipeline,and the final result was synthesized with tree composite strategy.Finally,a two-level LOD model was leveraged to optimize interactive experience.The experimental results show that the proposed algorithm can be well applied to large-scale unstructured grid data volume visualization,and all interaction delays are in milliseconds,which meets real-time interaction needs.