导向滤波优化的单幅去雾算法  

Algorithm for single image dehazing based on guided filtering optimization

作　　者：胡晨辉[1] 吕伟杰[1] 张飞[1] HU Chen-hui LU; Wei-jie ZHANG Fei(School of Electrical Engineering and Automation, Tianjin University, Tianjin 300072, Chin)

机构地区：[1]天津大学电气与自动化工程学院,天津300072

出　　处：《传感器与微系统》2017年第10期129-132,共4页Transducer and Microsystem Technologies

摘　　要：针对暗原色先验的单幅去雾算法计算复杂度高,无法满足交通监控系统中实时性需求,且大气光易受白色物体影响,以及天空区域易失真的缺陷和景物边界出现白边现象,提出了基于暗原色改进的快速去雾算法。采用四叉树搜索的算法对大气光进行精确估计,利用最大值滤波后的差值图像估计出天空区域,对透射率进行补偿,利用导向滤波器改进透射率并结合大气散射模型恢复无雾图像。实验结果表明:改进算法改善了原算法去雾效果的同时也提高了算法的速度。Aiming at problem that dark channel prior algorithm is computational complex and quite time consuming,can ＆#39;t meet the demand of real-time in traffic surveillance system,and atmospheric light is easily affected by the white object,the limitation of distortion happens in the sky area and also the Halo phenomenon in the scene boundaries,a new defogging algorithm is proposed based on guided filtering,which firstly uses quad-tree search algorithm to accurately estimate atmospheric light and then uses the difference image obtained by the maximum value filtering to compensate the transmission rate in the sky area. Combine atmosphere physical scattering model to recover the haze-off image. Experimental result proves that the improved algorithm improves dehazing effect,at the same time,increase rate of algorithm.

关 键 词：图像去雾 导向滤波 暗原色先验 四叉树搜索 最大值滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]