基于双重暗通道结合与高斯加权的去雾改进算法

针对暗通道先验算法去雾后图像存在伪影和景深突变处留有残雾的问题,提出一种基于双重暗通道结合与高斯加权的去雾方法.分别利用超像素块滤波和中值方块滤波求取两层暗通道,对两层暗通道进行像素级结合并构造高斯加权函数进行加强,然后采用引导滤波优化透射率图.把雾图转化到HSV空间并提取包含天空等的白色区域,引入概率函数取白色区域前10％像素点的明度分量平均值作为大气光估计值.实验结果表明,所提算法能较好地保留图像细节并去除景深突变处的残雾,同时能改善伪影现象.

Improved Image Dehazing Based on Combination of Double Dark Channels and Gaussian Weighting

To address the problem that the dark channel prior algorithm will produce artifacts and residual fog, a dehazing method was proposed based on double dark channel and Gaussian weighting, which use superpixel block filtering and median block filtering to obtain two layers of dark channels, respectively. Then a pixel-level fusion was performed on the two-layer dark channels and a Gaussian weighting function was constructed for enhancement. The white area that contains the sky was extracted based on the HSV color space and the average value of the brightness component of the top 10% of the pixels in the white area was taken as the atmospheric light. Experimental results demonstrate that the proposed algorithm can preserve details and remove residual fog while improving artifacts.