为了解决暗通道先验算法中对大片天空区域或大片白色区域失效的问题, 提出一种结合景深估计的高斯衰减与自适应补偿去雾算法。首先通过RGB通道散射强度的算术平均与雾浓度近似正相关的关系估计场景深度; 然后, 结合景深的边缘信息, 利用相邻像素之间的差异构造高斯滤波器对最小值通道进行滤波处理得到高斯暗通道; 其次, 利用高斯暗通道与其高斯函数之间的关系, 通过调节因子与雾浓度呈负相关的关系, 提出结合景深与高斯环绕函数卷积的策略获取自适应调节因子, 从而对透射率进行自适应补偿估计; 最后, 结合大气散射模型复原无雾图像。实验结果表明, 该算法在确保运行效率的基础上, 可以准确的估计透射率, 与经典算法相比较, 在客观评价中可见边数平均提高了0.02, 饱和像素点数平均下降了0.002。所提算法可以复原出自然清晰的无雾图像, 尤其是在景深远处和天空区域取得了良好的效果。

In order to solve the problem of large sky or white area failure in a dark channel prior algorithm, a Gaussian decay and adaptive compensation dehazing algorithm combined with scene depth estimation was proposed. Firstly, estimating the scene depth by an approximately positive correlation between the arithmetic mean of the RGB channel scattering intensities and the haze concentration. Then, combined with the edge information of the scene depth, a Gaussian filter is constructed using the difference between adjacent pixels to filter the minimum channel to obtain a Gaussian dark channel. Secondly, using the relationship between the Gaussian dark channel and its Gaussian function, through the relationship between the adjustment factor and the haze concentration, a strategy that combines convolution with scene depth and Gaussian surround function was proposed to obtain the adjustment factor; then, the transmission was adaptively compensated and estimated. Finally, the haze-free image was restored with the atmospheric scattering model. The experimental results show that the proposed algorithm can accurately estimate the transmission based on the operation efficiency. In the objective evaluation, the average number of edges increased by 0.02 while the number of saturated pixels decreased by 0.002. The proposed algorithm can also recover natural and clear haze-free images, especially in the scene depth and the sky area.