ReARM: a reconfigurable approximate Rounding-Based multiplier for image processing

Abstract

This paper aims to present a reconfigurable rounding based multiplier with different accuracy levels. It is based on a divide and conquer approach for applications in image processing. Our proposed approach divides the multiplicand and multiplier into two halves. Each half is multiplied with the other and our architecture is made accuracy-configurable. Further, due to rounding based approach, our approximate multiplication technique generates results faster. Based on our experiments, we observe that our proposed multiplier is 26.3% more accurate on an average compared to other state-of-the-art approximate multipliers for 8-bit operations