针对单幅图像去雾算法容易产生光晕现象且去雾后图像细节不突出的问题, 提出了一种基于各向异性扩散的去雾算法.首先在基于像素的暗通道先验假设的基础上计算出初始大气传输函数, 使用Perona-Malik偏微分方程模型求解出精细化的大气传输函数, 再经过最小值校正, 最终得到准确的大气传输函数。为了估计大气光, 对基于像素的亮通道图像进行像素排序, 从中选取出可靠的大气光向量。实验结果表明,提出的算法能够恢复更多的图像细节, 同时有效地抑制了光晕现象。

To suppress the halo effect of the existing dehaze algorithms and improve the detail of the recovered hazy image, a new single image dehazing algorithm based on anisotropic diffusion is proposed. Firstly, the transmission map is roughly estimated based on pixel-based dark channel prior, then the accurate transmission map is obtained by using Perona-Malik model and minimum pixel correction method. The atmospheric light is obtained from the sorted pixel-based bright channel pixels. Experimental results show that the algorithm can recover more details of the images and effectively suppress halo phenomenon.