Abstract:

Aiming at the problem that the foggy image restored by the dark channel prior algorithm produces color distortion and artifacts in the sky area, this paper proposes a new method that calculates dark channel confidence to correct the transmittance of the scene. The scene with higher saturation and lower brightness will more meet the hypothesis of dark channel prior. Combining the characteristics of saturation and brightness of foggy images, a dark channel confidence calculation method based on Gaussian function and adaptive parameter is designed to correct the transmittance of the scene. Experimental results show that the algorithm of the paper can effectively solve the problem of dark channel prior failure in the sky area, restore the image with natural colors, and there is no color distortion and artifacts.

Key words: image dehazing, dark channel prior, transmittance, atmospheric light value

摘要：

针对暗通道先验算法复原后的图像在天空区域容易产生色彩失真和伪影的问题，提出了一种新的暗通道置信度计算方法，用以校正天空区域的透射率。根据观察，图像中亮度越高、饱和度越低的区域，越不满足暗通道先验的假设，利用二维高斯函数，结合有雾图像饱和度和亮度的特点，设计了一种新的自适应调参的暗通道置信度计算方法，对不满足暗通道先验区域估算的透射率进行补偿。实验结果表明，该算法能有效解决暗通道先验在天空区域失效的问题，复原图像颜色自然，没有出现色彩失真和伪影。

关键词: 图像去雾, 暗通道先验, 透射率, 大气光值