A Novel Reconfigurable Computing Architecture for Image Signal Processing Using Circuit-Switched NoC and Synchronous Dataflow Model
In this paper, a novel reconfigurable architecture is proposed for multifunctional image signal processing systems. A circuit-switched NoC is used to provide interconnection because the non-TMD links ensure fixed throughput, which is a desirable behavior for computational intensive image processing algorithms compared with packet-switched NoC. Image processing algorithms are modeled as synchronous dataflow graphs which provide a unified model for general computing procedure. An image processing system is considered as several temporally mutually exclusive algorithms. Thus, their dataflow graph representations could be considered as a group and a merging algorithm could be applied to generate a union graph while eliminating spatial redundancy for area consumption optimization. After the union graph have been mapped and routed on the NoC, the reconfigurable system could be configured to any of its target image processing algorithms by properly setting the NoC topology. Experiments show the demo reconfigurable system with two image processing applications cost 26.4% less hardware resource, compared with the non-reconfigurable implementations.