结合自适应分割策略的图像去雾算法  

Single Image Dehazing Algorithm via Adaptive Threshold Segmentation

作　　者：王文杰 景志宇 梁晓瑞 潘婷 杨秋翔[1] WANG Wenjie;JING Zhiyu;LIANG Xiaorui;PAN Ting;YANG Qiuxiang(School of Software,North University of China,Taiyuan 030051;Key Laboratory of Instrument Science and Dynamic Testing,Ministry of Education,North University of China,Taiyuan 030051)

机构地区：[1]中北大学软件学院,太原030051 [2]中北大学仪器科学与动态测试教育部重点实验室,太原030051

出　　处：《计算机与数字工程》2022年第1期174-179,共6页Computer & Digital Engineering

基　　金：总装预研基金项目(编号:9140A17020113BQ04226);山西省高校科技基金项目(编号:20181560,20181562)资助。

摘　　要：暗通道先验去雾算法处理图像易出现局部失真问题,针对这一现象,提出一种单幅户外图像去雾算法。首先,利用四叉树分解的方法得到准确的大气光估计值;然后,结合暗通道图像的直方图分布特征,利用最大化类间方差策略自适应的估计暗通道图像分割阈值,并以此为先验知识优化透射率;最后,使用Gamma矫正提升图像整体对比度。实验表明,该算法能够有效地避免暗通道先验算法在天空区域失真的问题,对比其他算法,恢复图像视觉效果良好,客观评价指标均有所提升。The dark channel prior dehazing algorithm is used to process the outdoor image,the phenomenon of image distor⁃tion occurs.In view of this problem,an outdoor image dehazing algorithm is proposed.First,the more precise atmospheric light is obtained by quad-tree decomposition algorithm.Secondly,combining the image dark channel histogram distribution,the strategy of maximum class square is used to adaptively estimate the segmentation threshold of the dark channel image,and the transmittance is optimized by using this as prior knowledge.Finally,Gamma correction algorithm is used to improve contrast of the image.The exper⁃imental results show that the method can effectively avoid the image distortion phenomenon in sky area,comparing with other algo⁃rithm,both the visual effects and objective evaluation can be obtained the good results by the proposed algorithm.

关 键 词：四叉树分解 阈值分割 暗通道先验 最大类间方差 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]