Single Image Dehazing Using Dark Channel Prior and Adjacent Region Similarity

Abstract

Dark channel prior is an effective method in single image dehazing. However, it may work badly for the light color area in a haze image. In this paper, a novel method based on the adjacent region similarity is presented to solve this problem. The transmissions for the adjacent areas of light color region are estimated firstly. Then the similarity analysis of contrast between the light color area and the adjacent area is applied to help select the effective transmissions to estimate the transmission of the light color area. At last, the guided filter is used to refine the transmission map. Our method obtains a more accurate transmission map just using a single haze image. The final haze-free image is more natural and healthier. Robustness of the dark channel prior method is also strengthened.