一种基于梯度域引导滤波的图像去雾算法
An Image Defogging Algorithm Based on Gradient Domain Guided Filtering

利用暗通道先验理论进行单幅图像去雾时，通常采用最小值滤波算法估计透射率，该算法直接导致了图像边缘模糊、出现光晕等现象，或天空等含大面积雾霾区域时，出现颜色失真等问题。梯度域引导滤波在保持像素边缘方面的独特性能，结合针对天空区域去雾的分段颜色补偿方法，准确的解决了上述问题。改进算法与其他去雾算法进行了仿真比较，实验结果表明：改进算法可对图像景物和天空进行准确的去雾，与此同时该算法在颜色细节增强和高动态分布方面具有较大优势。
When using the prior theory of dark channel to remove fog from a single image, the minimum filtering algorithm is usually used to estimate the transmittance. This algorithm directly leads to some phenomena such as edge blur and halo, or color distortion in the sky and other regions with large area of haze. On the basis of studying the unique performance of gradient-domain guided fil-tering in keeping edge pixels, combined with the segmentation color compensation method for sky region defogging, the above problems are solved accurately. The improved algorithm was simulated and compared with other defogging algorithms. The experimental results demonstrate that the improved algorithm is capable of accurately defogging both the objects and the sky in the image. Additionally, the algorithm exhibits significant advantages in enhancing color details and handling high dynamic range distribution.