投影四面体法需要对四面体集合进行逐帧排序,而相邻四面体之间的顺序制约和依赖性限制了并行排序的效率.基于以上问题,提出一种视点相关的、基于逻辑切割的快速四面体集合精确排序方法.该方法分为4个串行步骤:首先沿视线方向将四面体集合所在空间剖分成一列有序的深度区间,每个区间包含一组四面体子集;然后并行地对每个区间的子集逐层提取互不遮挡的四面体,完成精确排序;再将区间边界上的四面体进行逻辑切割,并采用分段积分计算所有四面体在区间内的颜色贡献;最后按序沿视线方向累积所有区间的颜色贡献.实验结果表明,文中方法提高了四面体拓扑排序的并行度,极大地降低了排序时间,并大幅度地改进了绘制效率.
Projected tetrahedra needs to sort tetrahedral datasets by frame. Parallel performance has been limited by the sequence and the dependency of neighboring tetrahedral. Based on these questions, we propose a quick and accurate sorting algorithm of tetrahedral datasets, which is view-dependent and based on logical partition. The algorithm takes 4 sequential steps. Firstly, we divide the space containing tetrahedral datasets into a list of blocks along the viewing direction, with every block containing a subset of tetrahedra; then, we sort accurately by peeling tetrahedral subsets into unobstructed layers in parallel; thirdly, we partition tetrahedra on boundary logically and compute the optical contribution of each subset with segment integration; finally, optical contribution of each subset is accumulated along the viewing direction. The experiment results show that our algorithm enhances the parallelization of topological sorting, reduces the sorting time largely and improves rendering performance greatly.