Image Edge Extraction Using SOT-MRAM Based In-Memory Computing

Edge extraction is a crucial task in the realms of computer vision and image processing. Numerous algorithms for edge extraction traditionally depend on computationally intensive computations involving first-order or second-order derivatives, which, in turn, demand substantial energy consumption. Consequently, there is a need to explore other alternate methods that can be implemented in-memory and thereby reducing energy consumption. In this work, we explore a novel method of efficient edge detection using in-memory computing with spin-orbit torque magnetic random-access memory (SOT-MRAM). The paper presents a 4 × 1 SOT-MRAM array implementing XOR operations that can be used for extracting edges with 2 × 2 sub-matrices of a binary image followed by voltage comparisons. The proposed method targets low power consumption with minimal circuitry and shows a slight variation in performance metrics as compared to conventional approach of MRAM based edge detection operators such as Sobel, Canny Prewitt and Roberts.