Multilevel Image Dehazing Algorithm Using Conditional Generative Adversarial Networks

Kailei Gan,Hao Chen,Jieyu Zhao

doi:10.1109/access.2020.2981944

Kailei Gan, Hao Chen + Show 1 more

Open Access

https://doi.org/10.1109/access.2020.2981944

Copy DOI

Abstract

In recent years, the hazy weather in China occurs frequently, and image dehazing has gradually become a research hotspot. To improve the dehazing effect of the hazy images, this paper has proposed a multilevel image dehazing algorithm using conditional generative adversarial networks (CGAN). The hazy image is used to generate the composed image $K$ jointly estimated by a transmission map and atmospheric light value through a generator network, and a dehazed image is calculated through an improved atmospheric scattering model. The generator network and the joint discriminator network are subjected to adversarial training and reconstruction constraints. The experimental results show that the proposed method achieved good dehazing effect in synthetic hazy images and real hazy images, and is ahead of other advanced dehazing methods in subjective evaluation and objective evaluation.

Highlights

Hazy weather reduces the color saturation and contrast in images, and many image details are lost
Dehazed image is obtained by the improved atmospheric scattering model
EXPERIMENTAL RESULTS AND ANALYSIS This paper compares the synthetic hazy image test set and the real hazy image test set with a variety of advanced image dehazing methods to evaluate the performance

Summary

Introduction

Hazy weather reduces the color saturation and contrast in images, and many image details are lost. There have been many studies on image dehazing [1]–[6] He et al [7] proposed a dark channel prior method, which achieved good dehazing effects, but the soft matting method used had problems of low efficiency and a large number of calculations. He et al [8] proposed using guided filtering instead of soft matting, which increased the accuracy of transmission image estimation and improved the calculation efficiency. Ancuti and Ancuti [9] proposed a fusion-based method that fuses two images with enhanced contrast and white balance through a multiscale Laplacian pyramid. This can preserve the long-range and close-up information of the image, but it is possible

Methods

Results

Conclusion