Single Image Dehazing Algorithm Based on Sky Segmentation

Bad weather reduces the imaging quality of the intelligent vision system, such as haze and fog. Thus, haze removal has received wide attention from researchers. Most algorithms often suffer from color distortion and edge loss when dealing with the images containing large areas of sky. In this paper, we propose an effective dehazing method. The iterative threshold segmentation is used to segment the sky region out from the image, and then the brightness of the sky region is adjusted to increase clarity. The improved dark channel priori is used to process the rest regions. The transmission map is estimated by fast bilateral filtering. Finally, the two regions are merged together to get the haze removal result. Our algorithm achieves a clear and natural haze-free image, and has a faster processing speed. Meanwhile, it is universal and real-time in practical application.