Abstract: Due to scattering of atmospheric light and attenuation of incident light, the quality of images captured under foggy conditions will be reduced exponentially to the scene depth. In order to improve the visual effect of the degraded image, this paper makes use of the dark channel to estimate the air light. According to the strength of RGB channel, then, the scene depth map is estimated from a single fog image. Moreover, this paper introduces a tolerance threshold in the bright region to avoid the error estimates of the scene depth. A bilateral filter with edge-preserving is utilized to further process the scene depth map. A clear image can be recovered based on the atmospheric scattering model. Experiment results show that the proposed improved algorithm can enhance the visibility of fog-degraded images effectively.

Key words: atmospheric scattering model, RGB channel, dark channel, tolerance threshold

摘要： 因大气光的散射和入射光的衰减，大雾天气下捕获到的图像质量呈指数系数下降。为改善退化图像的视觉效果，利用暗原色先验信息获得大气光；根据单幅雾天图像RGB通道强度估计场景深度图；同时在明亮区域引入一种容差阈值，避免了这部分区域场景深度的错误估计，使用带边缘保持的双边滤波器对求得的场景深度图进一步处理；利用大气散射模型实现了图像复原。实验结果表明，改进后的算法有效提高了雾天退化图像的视觉效果。

关键词: 大气物理模型, RGB通道, 暗原色, 容差阈值