Modified Image Haze Removal Algorithm Based on Dark Channel Prior

Junpeng Hu, Zuoyong Li, Xinwei Chen

Abstract: Image defogging/dehazing is an important research hotspot in the field of image processing. The purpose of image defogging is to remove fog and obtain more image details. However, image defogging results based on dark channel prior usually have the problems of color distortion and insufficient brightness, which leads to the loss of partial image information. In order to solve this problem, we proposed a new image defogging algorithm based on dark channel prior. The proposed algorithm first selects the candidate regions of the sky by combining the location priori and brightness characteristics of the image pixel points, and reasonably estimate the atmospheric light value via the pixel points in the candidate regions. Second, the proposed algorithm respectively reduces the RGB (Red, Green, and Blue) channel values of the fog image, and combines each reduced channel and the other two previously unreduced channels to make up three new fog images. Finally, the proposed algorithm uses our image defogging algorithm to get three defogging images, and weights three defogging images to get the restored image. Comparative study and qualitative evaluation indicated that the proposed algorithm can obtain clearer defogging image.