摘要
传统的图像合成算法是为同构环境设计的。对于异构环境下顺序无关的图像合成问题,这些算法不考虑环境的异构性,也没有利用顺序无关合成操作符的性质,容易造成资源的浪费。针对这个问题,提出了一个异构环境下顺序无关的图像合成算法,它会根据环境的计算资源与网络资源更合理地进行任务分配,并利用顺序无关合成操作符的性质来优化合成方案。在ns-3网络模拟器上进行了大量的实验。实验结果表明:当环境存在较大异构性时,该方法相对经典算法binary swap能够取得显著的加速比。
Conventional image compositing algorithms are designed for homogeneous environments. These algorithms do not take heterogeneity into account and do not make use of the properties of an order-independent compositing operator. As for order-independent image compositing problems in a heterogeneous environment, these algorithms usually can not make full use of the resources. An order-independent image compositing algorithm in heterogeneous environments was proposed, which was able to distribute the workload according to computing and networking characteristics, and managed to optimize compositing scheme by utilizing the properties of order-independent operator. A comprehensive evaluation on the ns-3 network simulator was conducted. Experiment results show when the environment has great heterogeneity, the algorithm will gain significant speedup over classical algorithm binary swap.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2015年第9期2117-2125,共9页
Journal of System Simulation
基金
国家自然科学基金(61173067
61379085
61300131)
关键词
图像合成
异构环境
并行可视化
差分进化算法
ns-3网络模拟器
image compositing
heterogeneous environment
parallel visualization
differential evolution
ns-3 simulator
