去光源影响的加权导向滤波图像去雾算法  

Image Dehazing via Weighted Guided Filtering with Removing Lighting Effects

作　　者：毛德乾 高珊珊[1,2,3] 张晨昊 张彩明 Mao Deqian;Gao Shanshan;Zhang Chenhao;Zhang Caiming(School of Computer Science and Technology,Shandong University of Finance and Economics,Jinan 250014;Shandong Provincial Key Laboratory of Digital Media Technology,Jinan 250014;Shandong China-U.S.Digital Media International Cooperation Research Center,Jinan 250014;School of Software,Shandong University,Jinan 250101;Shandong Co-Innovation Center of Future Intelligent Computing,Yantai 264025)

机构地区：[1]山东财经大学计算机科学与技术学院,济南250014 [2]山东省数字媒体技术重点实验室,济南250014 [3]山东省中美数字媒体国际合作研究中心,济南250014 [4]山东大学软件学院,济南250101 [5]山东省未来智能计算协同创新中心,烟台264025

出　　处：《计算机辅助设计与图形学学报》2022年第2期217-231,共15页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(U1909210);山东省重点研发计划(2019GGX101007,2019GSF109112);山东省自然科学基金(ZR2020MF037,ZR2019MF016,ZR2019MF051);山东省高等学校青创人才引育计划和创新计划(2020KJN007);济南市科研带头人工作室(2021GXRC092).

摘　　要：含光源影响的雾图像中光源极易引入雾天原本不存在的光晕,从而影响大气光值和透射率估算的准确率,针对此问题,提出一种简单、易实现的去光源影响的雾天图像去雾算法.首先,基于超像素分割在CIELab颜色空间进行光源区域确定;然后,引入基于距离度量的光衰减因子计算并去除光源的影响;最后,以超像素块为单位估计大气光值,并采用加权导向滤波迭代优化透射率获得光源影响下的去雾结果图,提高效率和准确率.在光源影响下雾图集合和SOTS等数据集上进行了复原实验,将雾天图像复原结果与现有主流算法复原结果进行主客观对比,实验结果表明,文中算法复原结果的NIQE和BRISQUE值更低,并具有清晰度高、噪声低、纹理细节丰富和色彩恢复度高的优点.In fog images with the influence of light sources, the light source can easily introduce halo that does not exist in foggy days, which will affect the accuracy of atmospheric light value and transmittance estimation. To solve this problem, a simple and easy dehazing algorithm is proposed to dehaze fog images affected by light sources. Firstly, the light source area is determined in CIELab color space based on super pixel segmentation. Then the light attenuation factor based on distance measurement is introduced to eliminate the influence of light sources. Finally atmospheric light value is estimated in the unit of superpixel block, and weighted guidance filtering is used to iteratively optimize the transmittance to obtain the dehazing result map under the influence of light sources, which improves the efficiency and accuracy. The restoration experiment is carried out on the SOTS dataset and images affected by light sources. Comparing with existing mainstream algorithms subjectively and objectively, the restoration results of the proposed algorithm has lower NIQE and BRISQUE value. And these results have the advantages of high definition, low noise, rich texture details and high color restoration.

关 键 词：去雾 光源影响 光衰减 加权导向滤波 超像素 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]