基于黑色通道的图像快速去雾优化算法  

Optimized Fast Dehazing Method Based on Dark Channel Prior

作　　者：褚宏莉[1] 李元祥[1] 周则明[2] 沈霁[1] 

机构地区：[1]上海交通大学航空航天学院,上海200240 [2]解放军理工大学气象学院,江苏南京211101

出　　处：《电子学报》2013年第4期791-797,共7页Acta Electronica Sinica

基　　金：国家自然科学基金(No.41174164;No.61174196)

摘　　要：黑色通道优先算法在处理单幅户外场景图像去雾方面取得了一定的效果,但是该方法需要消耗大量的存储和计算资源,另外对天空区域的处理不够准确.本文仍然基于黑色通道先验知识,对图像边缘和非边缘分别采用不同的模板处理得到透射图,并通过分割天空区域或者雾最浓区域求得准确的大气光值,利用该方法可以恢复出和原方法基本一致甚至更准确的结果,并且可以大大减少消耗,提升计算速度.The dark channel prior algorithm is effective to some extent when working on some single outdoor scene hazing images. However, it has to expend large amounts of memory storage and computational time. Besides, the result of the sky region is not accurate enough. In this paper, we estimate the transmission map through processing the edges and non-edges respectively with different neighborhoods and obtain more accurate almosphefic light value by departing the sky part or the most hazed region of the image. This method can get a similar or even a better dehazing result than the classical haze removal methods, and moreover take less memory consumption and get much faster computational speed.

关 键 词：图像复原 黑色通道 去雾 天空区域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]