Nighttime Dehazing of Military UAVs Based on Joint Model Estimation of Adaptive Ambient Light and Saturation Line Prior

With the increasing demand for all-weather military UAV operations, nighttime image dehazing has become a critical yet challenging task due to uneven illumination, artificial light interference, and the inapplicability of daytime atmospheric scattering models. Traditional methods, such as dark channel prior (DCP), suffer from color distortion, overexposure, and poor robustness in low-light nighttime scenarios. To address these issues, this paper proposes a joint model estimation of adaptive ambient light and saturation line prior (SLP) for nighttime dehazing of military UAVs. Firstly, we use SLP to obtain more accurate transmission maps. Furthermore, adaptive ambient light is determined by fusing weighted maximum filtering for light-source regions and guided filtering for non-light-source regions. Finally, based on the calculated transmittance and adaptive ambient light, the haze-free image can be estimated according to the joint model. Experimental results demonstrate that the proposed method outperforms existing representative algorithms (DCP, SLP, ROP (rank-one prior), DehazeNet and DehazeFormer) in both qualitative and quantitative evaluations, which is specifically manifested in three aspects: (1) achieving superior haze removal effects; (2) enhancing image brightness and clarity preservation; (3) achieving stronger robustness in complex nighttime environments. This work provides a low-complexity, real-time, robust and effective solution for nighttime image dehazing, suitable for military UAV operations such as reconnaissance and target identification.