An Improved Single Image Haze Removal Algorithm Based on Dark Channel Prior and Histogram Specification

Abstract

Abstract. We introduce an improved single image haze removal algorithm, which combines dark channel prior (DCP) and histogram specification. First, the dark channel prior knowledge proposed by Kaiming He is analyzed and a conclusion is drawn that the haze removal image based on dark channel prior will have a tendency to dim and indistinct in some specific situations. Especially, when cleaning the haze in the image with large background area and low contrast, DCP result appears obvious anamorphose. Next, in order to improve the dehazing result of this kind of image, we propose an approach to change the contrast and intensity of haze removal image after DCP method by rebuilding the histogram of the image. Then, a modified approach is applied to fit general haze image. We experiment our method with a variety of outdoor haze images. The effectiveness of our method is demonstrated in comparison with DCP result when the input image contains low contrast scene and large background area, such as thick fog or dark surroundings in dusk. Our job makes up the deficiency of the dark channel model for this kind of image and enhance the contrast of the scene. Furthermore, the experimental results show that the dehazing effect on general haze image appears more close to real scene than dark channel model.

A review on dark channel prior based image dehazing algorithms

“…There are also several works considering the noise amplification problem during dehazed image construction [29,35]. In addition, to obtain high contrast dehazed images, some image processing techniques can be applied including as linear stretching [30], gamma correction [32], and histogram specification [33].…”