摘要
线积分卷积(LIC)是矢量场可视化中一个强有力的工具。但其计算量过大、耗时过多,影响了它的应用。根据LIC算法的特点,提出了LIC算法的并行实现方法。由于流线跟踪和卷积计算的独立性,只需把输出图像以像素点为单位平均分配给各处理器节点进行计算。从进程之间没有通信,仅当从进程开始计算前和计算结束后,在主进程与从进程之间有数据传递,通信开销很小。最后应用MPI在Linux集群环境下实现了该算法,实验结果表明,该方法具有较高的并行度和加速比。
Line integral convolution (LIC) is a powerful technique for vector field visualization. But LIC demands much running time, what limit its utilization. In order to enhance computation efficiency, presents a parallel algorithm for LIC depending on characteristic of LIC. Because the streamline tracing and convolution computing of every point is independent, all pixels of the output image are distributed to every node for being computed. No message delivers among the slave processes, and only a little of data delivers between a master process and the slave ones. The tests on the vector field visualization show that parallel algorithm has a good performance on Linux cluster.
出处
《计算机技术与发展》
2008年第12期22-24,共3页
Computer Technology and Development
基金
山东省科技厅资助项目(2007BS01002)
关键词
线积分卷积
矢量场可视化
并行计算
line integral convolution
vector field visualization
parallel computing