Underwater non-local adaptive descattering imaging
Jinghan Xu, Guojun Wu, Fei Feng, Xu Zhao, Бо Лю · Optics Express · 2025
Model-based underwater restoration offers high interpretability. However, particle scattering, glare, and uneven illumination often degrade the estimation of background light and transmittance. We propose an underwater non-local adaptive descattering imaging (UNADI) system that enables accurate parameter estimation for high-resolution restoration. The adaptive background light estimation mechanism of UNADI effectively mitigates interference challenges in complex underwater environments, while its multidimensional layer expansion resolves the fundamental incompatibility between conventional haze-line prior and monochrome data processing. Validated against local-based, non-local-based, deep learning, and polarization-based approaches, UNADI achieved 124% resolution improvement on USAF 1951 resolution targets. In 25 m line-scanning lidar missions, its 2.4× SNR boost under fourfold attenuation distances enables robust target reconstruction for underwater exploration.