基于暗原色的图像去雾改进算法研究  

Improved Image Dehazing Based on Dark Channel Prior

作　　者：马春波[1] 徐浪平 敖珺[1] 

机构地区：[1]桂林电子科技大学信息与通信学院,桂林541004

出　　处：《计算机与数字工程》2016年第1期118-122,共5页Computer & Digital Engineering

基　　金：国家自然基金(编号:61167006);广西自然科学然基金(编号:2012GXNSFBA053173)资助

摘　　要：何恺明提出的暗原色先验是一种简单有效的单幅图像去雾算法,很好地解决了雾天场景透射率的估算问题,图像复原效果清晰自然。但是原算法中的软抠图算法复杂度高、耗时长,对全天空区域和明亮区域处理效果不理想,恢复图像黯淡。针对这些问题,文章引入了引导图像滤波和自适应的容差机制,改进了原始算法。实验证明改进后的算法不仅大大降低了计算复杂度,同时也很好地保证了图像的去雾效果。He Kaiming proposes a very simple but powerful prior,called the dark channel prior,for single image haze for single image haze removal.The problem of estimation of transmission is solved well and the restored image is clear and natural.But the computational complexity of the soft matting is high and it is not well done to the sky region,the images after haze removal looks dim.To solve these problems,the guided image filtering,adaptive tolerance mechanism are introduced and the exposure is increased to improve the original algorithm.Experiments prove that the improved algorithm not only greatly reduces the computational complexity,but also ensure the effect of image haze removal.

关 键 词：图像处理 图像去雾 暗原色 引导图像滤波 容差机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]