Memristor-Based Approximate Adders for Efficient In-Memory Computing in Image Processing

This paper presents two novel memristor-based approximate adder designs using the V/R-R logic principle to enhance in-memory computing (IMC) efficiency while balancing accuracy. These adders leverage approximate computation to improve the performance tolerant of minor precision losses in applications such as image processing. Experimental and simulation results demonstrate significant improvements in computational speed and area efficiency over conventional designs. Integrated into a ripple carry adder (RCA), the proposed designs maintain acceptable image quality for image addition, with peak signal-to-noise ratio (PSNR) exceeding 30 dB in most cases.