Deep Learning‐Enhanced Underwater Imaging Using Random Lasers Based on Burr‐Like Ta ₂ O ₅ /Ag NPs

ABSTRACT Underwater optical imaging is often degraded by light scattering in chaotic media, leading to speckle noise, reduced resolution, and loss of detail. This paper presents a high‐resolution underwater imaging system that combines a Ta 2 O 5 /Ag composite (TOAC) structured random laser with the Real‐ESRGAN deep learning model. The TOAC structure prolongs photon multiple scattering paths through the scattering enhancement, and enhances the quantum yield of the gain through the localized surface plasmon resonance, which significantly reduces the random laser threshold and improves the quality factor to 1225.69. Compared with traditional lasers and halogen light sources, the image illuminated by this random laser has higher resolution and clearer details. Furthermore, random laser underwater imaging enhanced by Real‐ESRGAN deep learning can suppress speckles, boost image brightness, and recover details. For a resolution test chart imaged in turbid water, the application of Real‐ESRGAN nearly doubles the signal‐to‐noise ratio. Moreover, the reconstructed images of paramecia achieve a structural similarity index of 0.8532 against the undisturbed reference, representing a 2.44‐fold enhancement over the unreconstructed images. This study provides a new platform for high‐resolution and speckle‐free imaging in complex underwater environments, showcasing the potential of biological imaging, underwater biological monitoring, Underwater imaging, burr‐like Ta 2 O 5 /Ag NPs, deep learning, random laserand marine environmental research.