Single Image Defogging Algorithm Based on Sky Region Segmentation

Abstract

The visibility, contrast and color fidelity of images acquired under foggy weather will decrease to a certain extent, image defogging is an important part of image processing. Since the dark channel prior is not applicable to the sky area, which results in the color distortion and some noise of the restored fog-free image. Based on the segmentation of the sky region, it is proposed to separately estimate the atmospheric light value and transmission map according to different regions, and finally generates a defogging image using the optimized synthetic global map. In order to divide the sky domain and the non-sky domain more accurately, the two results obtained by the threshold segmentation method for the dark channel map and the improved threshold segmentation method for the original foggy image are considered at the same time. In the step of estimating the transmittance of the sky domain, the dark channel map of sky domain is obtained based on HSI color space, making the sky area more natural after defogging. Experimental results show that the algorithm can achieve good defogging effect, smooth edge transition between regions, and the defogging images are more natural and clear.