Comparative study of power reduction of various solutions of an image processing IP on modern FPGAs

In this paper, we describe the applicability and efficiency of several power reduction techniques applied on a modern 65nm FPGA. For the given algorithm, image erosion and dilation, two major solutions were tested and compared with respect to power and energy consumption. Firstly the algorithm was run on a general purpose processor (gpp) NIOS and then hardware architecture of an Intellectual Property (IP) was designed. Furthermore IPs design was improved by applying a number of power optimization techniques. They involved RTL level clock gating, power driven synthesis with fitting and appropriate coding style. Results show that hardware implementation is much more energy efficient than a general purpose processor and power optimization schemes can reduce the overall power consumption on an FPGA.