Abstract: Image compositing requires dense inter-node communications and could easily become the bottleneck of a parallel visualization system. Conventional image compositing algorithms assume the environment is homogeneous and take no heterogeneity into consideration. In a heterogeneous environment, this assumption is likely to cause unreasonable load distribution among the nodes. To address this problem, we present a group-based image compositing algorithm in heterogeneous environments. It is able to distribute the workloads among the nodes according to computing and networking characteristics of the heterogeneous environments. We conduct a comprehensive evaluation on the ns-3 network simulator. Experiment results shows that when the environment has great heterogeneity, our algorithm gains significant speedup over conventional methods.

Key words: image compositing, heterogeneous environment, parallel visualization, ns-3 simulator