摘要
针对目前存在的单幅图像去雾算法无法有效处理天空区域的缺点,提出天空识别的改进暗通道先验去雾算法.首先利用差分函数与暗通道信息结合的方式来去除天空区域中高亮像素带来的干扰,准确估计的大气光;随后改进暗通道先验中透射率的估计方式,使其能对天空区域和景物区域分别估计,得到初略的透射率图,再联合自适应中值滤波和双边滤波快速地修正细化;最后利用物理散射模型估计无雾的清晰图像.算法与多种存在的图像去雾算法相对比,结果表明算法不但能对图像中景物和天空进行准确的去雾处理,且运行时间还提高了至少29%,能有效用于监控系统中的图像去雾.
The single image dehazing methods can not deal with the sky regions effectively,therefore a newdehazing method with skyindentified improved dark channel prior has been proposed. Firstly,the variogram is combined with dark channel information to remove the interference of bright pixels in the sky regions and estimate the atmospheric light accurately. Then the dark channel prior is improved for measuring the rough transmission map of the sky region and scenery area respectively. And the joint filters by adaptive median filtering and bilateral filtering is used to optimize the transmission map for obtaining the refined transmission map. Finally,the clear image can be estimated by physical model of atmospheric scattering and optimal reflectance imaging. Comparing with the existing dehazing algorithms,the proposed method not only can handle both different scenes and sky regions,but also can obtain accurate dahazing results. Besides that,the running time has been improved at least 29%. The proposed method can be used in the practice of machine vision monitoring system.
作者
谭龙江
TAN Long-jiang(College of Economics & Finance,Haqiao University,Quanzhou 362021,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2019年第1期210-214,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(71571074)资助
华侨大学教改课题项目(15JGYB03)资助