曝光融合在单幅图像去雾中的应用  

Application of exposure fusion to single image dehazing

作　　者：陈晨[1] 胡士强[1] 张军[1] 

机构地区：[1]上海交通大学航空航天学院,上海200240

出　　处：《计算机应用》2012年第1期241-244,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61074106);航空科学基金联合资助项目(2009ZC57003)

摘　　要：针对雾天拍摄图像的降质现象,提出一种简单、有效的单幅图像去雾算法。首先利用暗原色先验知识估计出大气光亮度;然后根据雾天图像的成像物理模型,对每一像素的景深进行较高精度的亚采样,生成对应的虚拟无雾图像备选序列;最后,根据曝光融合算法提出的像素曝光评价指标,利用多分辨率形式的图像融合方法从备选序列中提取出清晰的无雾图像。实验结果表明,该算法既保证了复原图像清晰度,又具有较好的实时性。This paper proposed a simple and effective method to remove haze from a single input image degraded by bad weather.First,it estimated the airlight by using dark channel prior knowledge.Then,under a mathematical model for describing the formation of a hazy image,the depth of field of each pixel was sampled to get a virtual haze-free image sequence.Finally,an exposure criterion introduced by exposure fusion algorithm was used to extract a haze-free image from the image sequence by multi-scale fashion image fusion method.The experimental results show that the proposed method can yield good results and it is appropriate for real-time applications.

关 键 词：图像去雾 暗原色 像素评价 图像序列分析 曝光融合 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术]