基于局部维纳滤波的交通图像去雾霾算法  

The Traffic Image De-haze Algorithm Based on Locally Wiener Filtering

作　　者：肖杰[1] 李同亮[1] 翟东海[1] 

机构地区：[1]西南交通大学信息科学与技术学院,四川成都611756

出　　处：《成都信息工程学院学报》2014年第6期567-573,共7页Journal of Chengdu University of Information Technology

基　　金：国家自然科学基金资助项目(61461048);国家社会科学基金资助项目(12EF119);西藏自治区科技厅科技计划重点资助项目(Z2013B28G28/02);四川省科技创新苗子工程资助项目(20132010)

摘　　要：为了提高雾霾天退化交通图像的色彩保真度和清晰度,基于局部维纳滤波并结合暗通道先验理论提出一种快速去雾霾算法。传统的基于暗通道先验理论的去雾霾算法对单一户外场景图像取得了不错的效果,但在处理较高分辨率交通图像时,需要消耗大量的计算和存储资源;同时,不能很好处理图像中的一些白色区域,会导致该区域色彩失真。针对这两个问题展开研究：首先,基于局部维纳滤波并结合暗通道先验原则优化大气散射光,进而得到细化的透射率;其次,通过动态调节参数的大小以适应天空等明亮区域地去雾霾处理;最后,将所提算法用于实例验证,结果表明算法不但可以保证交通图像白色区域的色彩不失真,而且去雾霾效率提高4～25倍。In order to improve the color fidelity and clarity of the degraded image in haze days,this paper proposes a fast traffic image de-haze algorithm based on locally Wiener filtering and dark channel priori theory. Traditional image dehaze algorithm based on dark channel priori theory for outdoor scene images obtained good results,but when dealing with a high-resolution image,it needs to consume a large amount of computing and storage resources; meanwhile,it cannot perfectly deal with some white areas of the image which may cause color distortion in the region. Research is carried out on these two questions： First,based on local Wiener filtering and dark channel prior channel Atmospheric scattering light is obtained,and then the refined transmittance is gotten; second,By dynamically adjusting the size of the parameters,the algorithm perfectly deals with the haze of the sky and other white areas; Finally,the proposed algorithm is validated for instance,the results show that the algorithm can not only guarantee the color of the image of the white area without distortion,and can also increase the algorithm efficiency 4 to 25 times.

关 键 词：暗通道先验 雾霾图像 维纳滤波 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]