A Hazy Image Restoration Algorithm via JND Based Histogram Equalization and Weighted DCP Transmission Factor

Abstract

This paper proposed a three-step hazy image restoration method. Firstly, in HSI color space, we proposed a novel histogram equalization algorithm based on just noticeable difference (JND) to enhance the hazy image. Secondly, we apply the dark channel prior (DCP) to remove the image haze and compute the weighted transmission factor map. Finally, we integrated the enhancement results and dehazing results by the weighted transmission factor maps adaptively, and then the restoration result is obtained. Since the proposed method takes account of visual perception, and well balance image enhancement and image dehazing. Therefore, it results in a more natural haze-free scene for the human visual system (HVS). Experiment results demonstrate that our method achieves better restorations compared with some other state-of-the-art approaches.