Novel Image Dehazing Algorithm Using Scene Segmentation and Open Channel Model

Abstract

This paper presents a novel image dehazing method based upon scene depth segmentation and the open dark channel model. After analysis of the fog-day imaging model and the nature of the captured image fog distribution in the distant and near-field regions, the original image is divided into two areas, namely, the near and the far field. Subsequently, the transmittance in the far field is corrected further. Furthermore, the method uses an open-dark channel model and guided filtering to restore the image twice, so that the deblurred image boundary transition is more natural leading to the avoidance of the halo artifacts. The experimental results show that the proposed method recovers the clear fog shooting images better, the visual effects are significantly improved, and it has a wide domain of application.