All‐Optical Circle Fitting via Vortex‐Antivortex Convolution Merging Within a Single Metasurface

ABSTRACT Circle fitting is a fundamental operation in recognition, localization, and measurement tasks across diverse scientific and technical domains. Traditionally, this process relies on image acquisition and subsequent digital processing, typically implemented via integrated electronic circuits. However, the inherent latency from sequential post‐processing poses a significant challenge, particularly in applications that require real‐time responsiveness or involve high‐throughput data processing. Inspired by the Hough gradient method widely used in computer vision and leveraging the real‐time, post‐processing‐free nature of optical computing, we propose an all‐optical circle fitting technique based on a metasurface‐integrated 4f spatial filtering system. A pair of optical vortex–antivortex kernels are compactly merged within a single dielectric metasurface, allowing simultaneous gradient extraction and intersection accumulation via passive optical convolution. The proposed framework demonstrates high robustness, accurately detecting circular features across a broad range of radii and structural imperfections. This work establishes a platform for integrated optical shape analysis and underscores the promise of metasurface‐enabled optical computing in real‐time machine vision and advanced metrology.