MPEG-4 video consists of various video objects, rather than frames, allowing a true interactivity and manipulation of separate arbitrary shape object. Soware-based encoding of MPEG-4 video objects can be carried out by using parallel processing with efficient scheduling scheme to speedup the computation. In this paper, we propose two dynamic scheduling algorithms which have different scheduling costs and pegormance levels. The algorithms assign the multiple video objects encoding tasks to the cluster of workstations with proper load balancing. The algorithms allow user on-line interactions and pegorm the concurrent encoding on the video objects to achieve real-time speed. The experimental results, while showing real-time encoding rates, exhibit tradeoffs between load balancing, overhead scheduling cost and global pegormance.
Yong He, Ishfaq Ahmad, Ming L. Liou