在简要分析了当代几种典型的并行处理系统结构后,提出使用4颗TI公司高端数字信号处理器-TMS320C6416构建一种新型的并行图像处理系统。该系统通过一个同步四口SRAM和PCI总线构成互连结构,兼有紧耦合并行系统和松耦合并行系统的优点。然后,在所设计的并行硬件平台上,通过理论计算和实际的图像处理算法给出了包括点对点通信开销、加速比等在内的并行系统性能指标。最后,针对系统的性能瓶颈,进一步给出了提高系统性能的一些建议。 After a brief analysis of several typical contemporary parallel processing system architectures, a new parallel image processing system based on 4 high-end digital signal processors from Texas Instruments, TMS320C6416, is proposed. The system consists of a synchronous four-port SRAM and PCI bus to form an interconnect structure that combines the benefits of a tightly coupled parallel system with a loosely coupled parallel system. Then, on the designed parallel hardware platform, the performance indexes of parallel system including point-to-point communication overhead and speedup are given through theoretical calculation and actual image processing algorithms. Finally, according to the performance bottleneck of the system, some suggestions for improving system performance are given further.