Improved Dark Channel Prior Dehazing Algorithm Combined with Sky Recognition

A dark channel a prior defogging improvement algorithm combined with sky recognition is proposed to address the problems of failure of dark channel a prior for sky regions, artifacts and detail loss. Firstly, a quadtree search algorithm is used to find the atmospheric light values; then, an adaptive weighted guide filter is used to improve the transmittance map to enhance the edge details; the sky region is segmented by setting the brightness threshold and sky features, and the lower limit of transmittance is corrected according to the results of sky region identification; finally, the solution of the recovered clear image is carried out by substituting the atmospheric scattering model. The experiments show that the improved algorithm can improve the problems of image detail loss, halo, appearing artifacts, etc. compared with other defogging algorithms, and effectively solve the phenomenon of failure to sky regions. Compared with the classical dark channel a prior algorithm, the algorithm in this paper improves 37.02% in PSNR and 47.56% in SSIM.