一种改进的有雾图像去雾算法  

An Improved Foggy Image Dehazing Algorithm

作　　者：何涛 谢有庆 李一航 徐昊 刘甲甲 余萍[3] HE Tao;XIE You-qing;LI Yi-hang;XU Hao;LIU Jia-jia;YU Ping(Shenzhen Nanshan Power Supply Bureau Co.,Ltd.,Shenzhen 518000,China;Shenzhen Power Supply Planning and Design Institute Co.,Ltd.,Shenzhen 518000,China;Department of Electronic and Communication Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]深圳南山供电局,深圳518000 [2]深圳供电规划设计院有限公司,深圳518000 [3]华北电力大学电子与通信工程系,保定071003

出　　处：《科学技术与工程》2020年第2期675-680,共6页Science Technology and Engineering

基　　金：深圳供电局有限公司项目(090000KK52180035)。

摘　　要：针对基于大气散射模型的图像去雾算法存在的图像去雾后颜色偏暗、对比度过度增强的问题,提出一种基于灰色关联度引导滤波的图像去雾算法。首先,假设有雾图像的像素可以分为正常像素和被雾霾颗粒破坏的像素,应用灰色关联理论对雾霾图像的像素值进行判断;然后,对雾霾颗粒破坏的像素进行引导滤波,通过取对数的方法缩小原始图像和滤波以后图像像素值之间的差异,在对数域中计算雾气面纱值;最后,依据大气散射退化模型反演复原清晰的图像。实验结果表明,该算法不仅可以有效改善雾霾图像的清晰度,而且能够解决去雾后存在的亮度偏暗,色彩失真等问题。Image defogging in atmospheric scattering model often results in dimmed or over-contrasted images processed.To solve the problem,a fast image-defogging algorithm based on gray-correlation-degree-guided filtering was proposed.First,assuming that the pixels of foggy images could be divided into normal pixels and haze-affected ones,the gray correlation theory was applied to judge the pixel values of hazed images.Second,the haze-affected pixels were guided to a filter,by which the difference in pixel value between the original image and the filtered image could be reduced after logarithmization of the values,and the fog veil values be calculated in logarithm domain.At last,a clear image could be retrieved in the atmospheric scattering degradation model.The experimental results show that the proposed algorithm can improve effectively the clarity of hazed images,and in addition,solve the problems of dim brightness and color distortion after defogging.

关 键 词：大气散射 灰色关联 引导滤波 对数域 大气面纱 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]