Improvement of Dark Channel Defogging Algorithm Based on Duided Filtering

Abstract

Aiming at the problem that the dark channel prior defogging algorithm has the problems that the dark channel fails in large areas such as sky or white objects and the guided filtering algorithm is not completely defogging, a single image defogging improvement based on dark channel prior and guided filtering is proposed algorithm. First obtain the minimum image of each color channel of the foggy image and the rough estimate of the transmittance according to the dark channel theory, use the guide filter for the minimum image to obtain the guide image containing the edge information, and then use the guide filter again to optimize the initial transmittance. The problem of underestimation of the transmittance of the foggy sky area is to determine the range of the sky area and make certain adjustments to its transmittance, and use the quadtree search algorithm to obtain accurate atmospheric light values. Finally, the fog-free image is recovered through the atmospheric physical model, and the algorithm is compared with other defogging algorithms. The results show that the algorithm can effectively process foggy images containing sky areas, and the time complexity is low.