End-to-end edge detection on self-rectifying resistive memory array

Abstract

In recent years, memristor that can integrate storage and computing has gained some attention in image processing. In this study, edge detection for noisy images is realized in a single-step manner by a new Gauss–Laplace operator in a circuit with a memristor array. It ensures the edge detection end-to-end and reduces hardware buffer consumption compared to the traditional two-step methods. Conductance values from our experimentally fabricated Cu₂Te memory device are used in the circuit design, which mitigates the gap between theory and real-world device arrays. Furthermore, an interesting finding is that the edge detection results in the circuit are closer to the ideal ones after reducing the grayscale from 256 to 16, which means fewer input voltage levels and a simpler input circuit are possible. It is hoped this work can help design memristor array circuits for image processing.