Scene-driven metasurface infrared super-resolution imaging

Abstract

Metasurfaces have attracted considerable attention in infrared imaging owing to their ultrathin structure, lightweight form, and integration potential. However, metasurface infrared imaging systems strongly rely on accurately calibrated imaging models and are vulnerable to detector nonuniformity, vignetting and infrared glare, spatially varying point spread functions, and undersampling in real dynamic scenes. To address this problem, this paper proposes a scene-driven self-calibration framework for metasurface infrared super-resolution imaging. By exploiting redundant information from natural motion, the proposed framework continuously updates and adaptively optimizes the imaging model without additional calibration data or true high-resolution ground truth. It combines nonuniformity correction, vignetting and glare suppression, multi-frame super-resolution reconstruction, and spatially varying PSF deconvolution to recover high-resolution images. Experimental results show that the proposed method achieves 4× pixel super-resolution and effectively improves the resolvability and contrast of stripe structures, providing a new approach for ultracompact infrared imaging systems.