Image dehazing using non-symmetry and anti-packing model based on dark channel prior

As is known to all, image processing is increasingly important in daily life and research. Image enhancing and image restoration are of vital importance in image processing. During the last decade, fog removal-— which is related to image enhancing and image restoration — received lots of attention. Some famous methods like dark channel prior were proposed. In this paper, a novel method based on the Non-symmetry and Anti-packing Model (NAM) is proposed to improve the dark channel prior algorithm. Besides, to enhance haze-free image's visual effects, auto level is used in this paper. Firstly, according to the basic theory of dark channel prior, which is based on a key observation that most local patches in outdoor haze-free images contain some pixels whose intensity is very low in at least one color channel, we can get several equations to calculate haze-free image. Then, we use the NAM to calculate an important constant value which is called atmospheric light. Next, we make use of a guided filter to estimate the transmission (t(x)) in a more accurate and subtle way. Finally, we can use estimated transmission and atmospheric light, which are refined due to the NAM and the guided filter to obtain the haze-free image, but the final haze-free image always has low light intensity, so we propose a method named auto level to enhance image's visual effects. According to the experiments, our proposed algorithm contributes to dehazing and it is efficient in fog removal.