Let You See in Sand Dust Weather: A Method Based on Halo-Reduced Dark Channel Prior Dehazing for Sand-Dust Image Enhancement

Zhenghao Shi,Erhu Zhang,Yaning Feng,Lifeng He,Minghua Zhao

doi:10.1109/access.2019.2936444

Abstract

The images that are captured in sand storms often suffer from low contrast and serious color cast that are caused by sand dust, and these issues will have significant negative effects on the performance of an outdoor computer vision system. To address these problems, a method based on halo-reduced dark channel prior (DCP) dehazing for sand dust image enhancement is proposed in this paper. It includes three components in sequence: color correction in the LAB color space based on gray world theory, dust removal using a halo-reduced DCP dehazing method, and contrast stretching in the LAB color space using a Gamma function improved contrast limited adaptive histogram equalization (CLAHE), in which a guided filter is used to improve the artifacts of the histogram equalization. Experiments on a large number of real sand dust images demonstrate that the proposed method can well remove the overall yellowing tone and dust haze effect and obtain normal visual colors and a detailed clear image.

Highlights

Images that are captured during sand storms usually suffer from low contrast, noise, and color distortions due to the scattering and absorption of light rays by dust particles
Aiming at problems that were mentioned above, we develop an improved dehazing method based on the dark channel prior for sand dust degraded image enhancement
In this paper, a method based on halo-reduced dark channel prior (DCP) dehazing for sand dust image enhancement is proposed

Summary

INTRODUCTION

Images that are captured during sand storms usually suffer from low contrast, noise, and color distortions due to the scattering and absorption of light rays by dust particles. SAND DUST IMAGE CONTRAST ENHANCEMENT the optical imaging process is more intricate in a sand dust environment than a hazy environment, sand dust images are similar to hazy images in some optical characteristics, such as atmospheric scattering and absorption Considering this point, many researchers applied dehazing methods for sand dust image enhancement, including histogram equalization [1]–[6], Retinex methods [7], the Laplacian filter [8], Nonlinear functions such as the Gamma function [9], [10], various image prior (such as the dark channel prior (DCP), color attenuation prior, and sparse prior)-based methods [11]–[17], [34], and various deep learning-based methods [18]–[20]. Aiming at problems that were mentioned above, we develop an improved dehazing method based on the dark channel prior for sand dust degraded image enhancement . Assuming that Rav, Gav, and Bav respectively represent the average value of the R, G, and B color channels in an image, and Rc, Gc, and Bc respectively represent the corrected values of the R, G, and B color channels, we have the following: Rc

B Gav Bav

ILLUMINATION ADJUSTMENT

DATA SET AND PARAMETER SETTING

DISCUSSIONS

CONCLUSIONS