Toward Energy‐Efficient Machine Vision: Advances in Optoelectronic Memristors
Shuaiwen Pan, Siqi Wu, Jianyu Ming, Haifeng Ling
Abstract
Abstract The growing demands of real‐time visual processing in edge intelligence necessitate innovative hardware paradigms that overcome the limitations of conventional vision systems. Optoelectronic memristors (OEMs), which intrinsically combine light sensitivity with tunable nonvolatile resistance states, have emerged as promising building blocks for energy‐efficient neuromorphic vision architectures. By enabling in‐sensor computing through photonic–electronic coupling, OEMs support the direct integration of sensing, memory, and computation within individual devices. This review presents a comprehensive overview of recent progress in OEM research, encompassing operational mechanisms, material platforms, functional emulation, and application scenarios. Remarkable progress is made in areas such as wavelength‐selective response, weak‐light perception, and negative photoconductivity (NPC), facilitating its deployment in static and dynamic image recognition as well as adaptive visual processing. The remaining challenges, such as material stability, response speed, and large‐scale array integration, are also examined alongside perspectives on future development toward scalable, high‐throughput, and energy‐efficient machine vision systems.