Image dehazing network based on depth information and bilateral grids
Yuchen Zhang, Chao Sun, Hanren Wang, Yueyue Liu, Qi Sun, Haonan Ni, Long Peng, Xinnan Fan · 2025
Images captured in hazy weather are susceptible to the natural environment, with significant problems such as decreased contrast and reduced visibility. Traditional studies generally use the priori knowledge contained in the image. However, due to the complexity of the real environment, the apriori-based method is relatively limited and can not achieve stable image dehazing. Recently, deep learning have been applied in numerous computer vision problems. The learning-based method uses a large amount of data for training and has stronger robustness. However, current image dehazing methods do not effectively utilize the depth information contained in the image and suffer from insufficient image edge information, resulting in incomplete dehazing and blurred details. To address these problems, we propose a novel image dehazing network based on depth information and bilateral grid. The integration of depth information makes the network better handle the uneven haze in the image. The bilateral grid contains spatial information and pixel value information, which can preserve the edge structure of the image. Experimental results on public datasets show that our proposed network achieves superior performance over the current methods.