Metaoptics merging computational optics and optical computing toward intelligent visual perception

Visual perception systems are evolving toward multifunctionality, miniaturization, and real-time intelligence, necessitating coadvancements in optical hardware and computational software. Computational optics enhances optical encoding through algorithms but faces computational limits, while optical computing offers photonic parallelism yet struggles with hardware complexity. Metaoptics, using subwavelength nanostructures for multidimensional light-field control, resolves this tension by serving dual roles: as an efficient computational optics encoder offloading algorithmic burdens and as a nanoscale optical computing processor enabling ultrahigh-density parallelism. This synergy drives computational optical integration for next-generation optoelectronics. This review outlines computational optics and optical computing fundamentals, analyzes metaoptics' advantages in multidimensional optical information encoding and processing, and details recent metaoptics-based advances and challenges in computational optics, optical computing, and their convergence. The conclusion provides a future roadmap, highlighting metaoptics-driven integration's potential for chip-scale intelligent visual perception by bridging the hardware-software divide to enable miniaturized devices with real-time adaptability, massive parallelism, and energy efficiency.