A real-time image dehazing method based on image restoration and enhancement techniques
Peiqi Wang, Jihao Cai, Hui Tian, Yangjie Cao, Yiduo Mei · 2025
Foggy environments significantly reduce visibility, leading to degraded image quality and loss of detail in imaging devices. This affects the clarity of edge surveillance footage and the performance of real-time computer vision tasks. Although dehazing algorithms have made significant progress in various fields, most methods primarily focus on synthetic hazy images, neglecting their performance on real hazy images, resulting in issues such as color distortion and haze residue. Furthermore, existing methods often incur high computational costs, making them unsuitable for real-time dehazing in monitoring scenarios. To address these challenges, we proposed a dehazing method based on image restoration and enhancement techniques. The method integrates an RDNet designed with residual networks and attention mechanisms for image restoration, and introduces a color enhancement module to improve image quality. By using a novel composite loss function, the performance and dehazing effectiveness of RDNet are significantly improved. Experimental results show that the proposed method outperforms most models in dehazing both synthetic and real hazy images, while significantly enhancing image quality with low computational cost. This method is of great practical value for real-time dehazing in edge surveillance devices under foggy conditions.