Hybrid Metalens-Compressive Sensing Imaging System for Ultra-Low-Power Edge Detection.

Low-power edge detection is essential for next-generation vision systems that require real-time on-device processing of optical information. Conventional imagers rely on power-intensive digital postprocessing for edge detection, creating a data bottleneck between the image sensor and signal processor. We propose a system-level hybrid optoelectronic edge imaging framework that combines an analog optical computing metalens with a compressive sensing CMOS image sensor (CS-CIS). The spiral metalens with a topological charge of l = + 2 directly images and encodes second-order edge information in the optical domain, eliminating the need for additional bulky optics and digital edge extraction algorithms. This optically processed information is captured by the CS-CIS that selects a reduced number of pixel values, minimizing redundant data acquisition and enabling high-speed and ultra-low-energy operation. Simulation results validate the effectiveness of this approach, demonstrating a peak signal-to-noise ratio of 23.87 dB at a 25.0% compression rate. This work paves the way for energy-efficient edge-aware systems with potential applications in robotics and computer vision.