MAGNet融合导向滤波的真实图像去雾方法  

Real-world Image Dehazing Method of MAGNet Fusing Guided Filtering

作　　者：桑榆 申红倩 张世辉[1,2] 路佳琪 左东旭 牛景春 SANG Yu;SHEN Hong-qian;ZHANG Shi-hui;LU Jia-qi;ZUO Dong-xu;NIU Jing-chun(School of Information Science and Engineering,Yanshan University,Qinhuangdao,Hebei 066004,China;The Key Laboratory for Computer Virtual Technology and System Integration of Hebei Province,Qinhuangdao,Hebei 066004,China)

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]河北省计算机虚拟技术与系统集成重点实验室,河北秦皇岛066004

出　　处：《计量学报》2022年第3期346-354,共9页Acta Metrologica Sinica

基　　金：中央引导地方科技发展资金(216Z0301G);河北省自然科学基金(F2019203285)。

摘　　要：为了提高单幅图像去雾方法的准确性以及适用范围,提出一种MAGNet融合导向滤波的真实图像去雾方法。首先,根据雾在真实图像中分布特性以及成像原理,设计多注意力残差密集块,从而有效提取真实图像中与雾相关的特征并降低梯度消失风险;其次,构造基于所设计的多注意力残差密集块的端到端卷积神经网络,实现对有雾图像中雾的去除;最后,将导向滤波引入去雾问题中,实现对去雾后真实图像视觉效果的增强。实验结果表明:与已有代表性的图像去雾方法相比,该方法不仅能够对真实图像进行去雾,还可以对合成图像中的雾进行有效去除,且去雾效果更佳。In order to improve the accuracy and applicable scope of the single image dehazing method,a real-world image dehazing method combining multiple attention grid network(MAGNet)and guided filtering is proposed.Firstly,according to the distribution characteristics and imaging principles of haze in real-world images,the multi-attention residual dense block is designed to effectively extract haze-related features in real-world images and the risk of gradient disappearance is reduced.Secondly,the end-to-end convolutional neural network based on the designed multi-attention residual dense block is constructed to remove haze from hazy images.Finally,the guided filtering is introduced into the dehazing problem to enhance the visual effect of dehazing images.The experimental results show that compared with the existing representative image dehazing methods,the proposed method can not only remove haze from real-world images,but also effectively remove haze from synthetic images,and the visual effect of dehazing image is better.

关 键 词：计量学 MAGNet融合导向滤波 真实图像去雾 多注意力机制 图像处理 单幅图像去雾法 

分 类 号：TB96[机械工程—光学工程]