Single image haze removal for aqueous vapour regions based on optimal correction of dark channel

Haze removal is an interesting topic in multimedia and image processing for many applications. Specially for the automatic piloting of ships, the haze removal technology for aqueous vapour regions plays a key role in safe piloting. However, the existing haze removal methods did not dehaze well for these areas. Based on this motive, this paper presents a new haze removal approach to improve the dehazing effect for aqueous vapour regions, in which we design two new computing mechanisms. The first one is to propose a new gradient change model of the dark channel value related to aqueous vapour regions. The second one is to design an optimized and iterated correction method for the dark channel of aqueous vapour regions. Finally, based on these two computing mechanisms, a dynamic iterative optimal correction model is presented to solve the proposed method. Both the visual and the quantitative experiments demonstrate the proposed method outperforms both the family methods of dark channel prior and the deep learning-based methods in aqueous vapour regions. In conclusion, the proposed method can effectively remove the haze in aqueous vapour regions.