改进BSP模型在实时图像并行处理中的应用

摘要

Considering the real-time processing on large numbers of image data, an improvement on BSP computational model is made. Based on it, several issues have been settled such as the division of super-step, the loss of super-step, the transmission of data, etc. At the same time, the cluster architecture adapting to paralell computing for the real-time images has been designed. Meanwhile, using the radio-style communication greatly reduces the time of communication and improves capability of the real time. This article analyzes the performance of the parallel computing in the fields of the speedup-ratio and efficiency. The validity of this model has been verified by experiments.%针对大数据量图像处理的实时性,改进了BSP计算模型,解决了超步划分、超步丢失、数据传输等问题.设计了适合实时图像并行处理的集群结构.采用广播式的通信方式极大地缩短了通信时间,提高了实时性.从加速比、效率方面分析了并行计算的性能,实验证明了此方法的有效性.