HardGAN: A haze aware representation distillation GAN for single image dehazing

In this paper, we present a Haze-Aware Representation Distillation Generative Adversarial Network (HardGAN) for single-image dehazing. Unlike previous studies that intend to model the transmission map and global atmospheric light jointly to restore a clear image, we approach this restoration problem by using a multi-scale structure neural network composed of our proposed haze-aware representation distillation layers. Moreover, we re-introduce to utilize the normalization layer skillfully instead of stacking with the convolutional layers directly as before to avoid useful information wash away, as claimed in many image quality enhancement studies. Extensive experiments on several synthetic benchmark datasets as well as the NTIRE 2020 real-world images show our proposed HardGAN performs favorably against the state-of-the-art methods in terms of PSNR, SSIM, LPIPS, and individual subjective evaluation.