基于图论分割的图像去雾算法  

Improved Haze Removal Algorithm Based on Graph-cut

作　　者：李娜娜 谭靖 

机构地区：[1]武汉航天远景科技股份有限公司,湖北武汉430200 [2]北京航天泰坦科技股份有限公司,北京100071

出　　处：《地理空间信息》2017年第8期80-82,共3页Geospatial Information

基　　金：国家水污染控制与治理科技重大资助项目(2012ZX075010020009);北京市科技新星计划资助项目(Z131101000413086)

摘　　要：雾天等恶劣天气条件下拍摄到的户外图像质量下降,严重影响视觉监控系统的可靠性和鲁棒性。为提高降质图像的对比度和清晰度,考虑到同一幅图像中不同景物的场景深度不同,采用基于图论的图像分割方法对图像进行分割,然后在分割后的各个区域,根据暗原色先验理论估算图像的尘雾浓度图,再根据DOS原理,将尘雾浓度图从原始图像中去除,得到复原后的清晰图像。实验表明,所提出的算法可以有效改善降质图像的质量,提高图像的对比度和清晰度。The images captured in haze are poor visually,which seriously affected the reliability and robust of visual surveillance system.In order to improve the contrast and sharpness of images,considering that different scenes were different in depth,we used graph-cut to segment images.After that,we obtained the fog concentration diagrams in each area based on dark-channel prior method.Finally,according to DOS theory,we removed the fog concentration diagram from the original image.With a series of experiments,we concluded that the improved algorithm is effective in improving the quality,contrast and sharpness of images.

关 键 词：图论 图像分割 去雾 DOS 

分 类 号：P231[天文地球—摄影测量与遥感]