摘要
针对现有的单幅图像去雾问题,提出了一种基于并联卷积神经网络的单幅图像去雾算法,以端对端的方式实现图像去雾.首先,使用雾天RGB图像YUV变换的Y、U和V分量构建并联卷积神经网络,自适应获得雾霾特征;网络结构由两个子网络组成,较深的网络预测清晰图像的亮度通道,较浅的网络预测色度通道和饱和度通道.最后,采用递归双边滤波,对去雾后的图像进行滤波,可以得到更加清晰的无雾图像.实验结果表明,本文去雾算法无论是在合成雾天图像数据集还是自然雾天图像数据集上,都具有良好的对比度与清晰度.在主观评价和客观评价方面,本文去雾算法都优于其他对比算法.
Aiming at the problem that the existing single image dehazing algorithm,a single image dehazing al-gorithm based on multiple convolutional neural networks is proposed.Firstly,the Y,U and V components trans-formed by YUV of foggy day RGB images were used to construct a multiple convolutional neural network to ob-tain haze characteristics adaptively.The network structure is composed of two subnetworks,the deeper one pre-dicts the brightness channel of the clear image,and the lighter one predicts the chromaticity channel and satura-tion channel.Finally,recursive bilateral filtering is adopted to filter the image after dehazing to obtain a clearer fog-free image.The experimental results show that this algorithm has good contrast and clarity in both synthetic and natural foggy image data sets,and is superior to other comparison algorithms in terms of subjective and objective evaluation.
作者
陈清江
张雪
CHEN Qing-Jiang;ZHANG Xue(College of Science,Xi'an University of Architecture and Technology,Xi'an 710055)
出处
《自动化学报》
EI
CAS
CSCD
北大核心
2021年第7期1739-1748,共10页
Acta Automatica Sinica
基金
国家自然科学基金(61403298)
陕西省自然科学基金(2015JM1024)
陕西省教育厅专项科研计划(2013JK0586)资助。
关键词
图像去雾
卷积神经网络
大气散射模型
多尺度卷积
递归双边滤波
Image dehazing
convolution neural network
atmospheric scattering model
multi-scale convolution
re-cursive bilateral filtering