A dark channel prior dehazing algorithm based on adaptive window filtering and dual channel weighted atmospheric light value estimation

Dark channel prior (DCP)-based image dehazing algorithms often rely on fixed window size filtering, which can result in block artifacts, partial color distortion, overall darkness, and high saturation in the dehazed image. To overcome these issues, this study introduces a dark channel prior dehazing algorithm based on adaptive window filtering and weighted dual channel atmospheric light value estimation (AF&DCE DCP) algorithm. This approach aims to enhance dehazing accuracy and image naturalness. The proposed method dynamically adjusts the filter window size based on the gradient values and applies a linear weighting technique to estimate atmospheric light using bright and dark dual-channel images, ensuring more precise estimations and improved dehazing results. Experimental comparisons with various related DCP algorithms demonstrate that the proposed method significantly improves multiple metrics, such as SSIM, RMSE, and information entropy (IE), indicating its ability to enhance image quality and robustness effectively.