Atmospheric light veil correction dehazing algorithm based on regional detection

Abstract

To address the problems that the dehazing algorithm is prone to distortion, block artifacts, and edge pixel loss from the minimum filter in the highlighted area of an image, a dehazing algorithm with an atmospheric veil correction model under regional detection is proposed. First, the initial atmospheric light veil is obtained by using the improved dark channel, and the atmospheric light veil correction model is established for it. Then, the color distance and color distance histogram fitting curve of the haze image are used to accurately detect the region where the initial atmospheric light veil is corrected, and the correction term is established. Finally, the atmospheric light is estimated by combining the image information entropy with quad-tree subdivision algorithm. It is proved that the proposed method effectively solves the distorted problem in the highlighted region and the block artifacts produced by minimum filtering and can recover image with rich detailed information.