Single Image Defogging Method Based on Improved Generative Adversarial Network
Yunfei Li, Jixiang Cheng, Zhidan Li, Qiwei Pan, Rui Zeng, Tian Tian · 2022 IEEE Asia-Pacific Conference on Image Processing, Electronics and Computers (IPEC) · 2022
Aiming at the problems of blurred image content, reduced contrast, color distortion, loss of information, and degradation of scene images caused by haze weather, an improved end-to-end single image defogging method is proposed. This method can learn transmission map, atmospheric light value and dehazing at the same time, including two stages: physical stage and deep learning stage. Then two kinds of modules are introduced in the second stage, including; attention mechanism and inception. The experimental results show that compared with the benchmark method of this article, the dehazing rate of psnr and ssim in the indoor and outdoor data sets are improved by 0.1 and 2.0 points, respectively. To mention the effectiveness of the method.