Design of Efficient Approximate Multiplier for Image Processing Applications

TLDR: A new truncation scheme and an error correction term which are applied to recursive multiplier architecture and prove that the proposed multiplier performs better compared to the existing multipliers.

Abstract: Approximate computing is an emerging paradigm to create energy-efficient computing systems. Most of the image processing applications are inherently error-resilient and can tolerate the error up to a certain limit. In such applications, energy can be saved by pruning the data path modules such as a multiplier. In this paper, we propose a new truncation scheme and an error correction term which are applied to recursive multiplier architecture. Further, truncation method and correction term that compensates the error in the proposed approximate multiplier significantly reduce the area, delay and power. Finally, the proposed multiplier is validated on an image sharpening algorithm. Simulations carried out clearly prove that the proposed multiplier performs better compared to the existing multipliers.