A Novel Single Image Dehazing Method

In this paper, a novel single image dehazing method based on dark channel prior is proposed. The original dark channel prior based method employs soft matting algorithm to refine the transmission map, which is time-consuming. Recently, guided filter, which is time-saving, has been introduced to perform the same function as soft matting. In our work, based on the observation that haze mainly has an effect on the low-frequency component of an image, by employing guided filter instead of soft matting algorithm to refine the transmission map of the low-frequency component of haze image extracted by Haar wavelet transform, the runtime cost is dramatically decreased. Experimental results show that the proposed algorithm restores the image faster while maintaining relatively excellent performance compared with the Dark Channel Prior with Guided Filter based method, which allows this algorithm to be applied to the real-time applications such as obstacle detection and surveillance.