DLL-GAN: Degradation-level-based learnable adversarial loss for image enhancement

Mohamed Kas, Abderrazak Chahi, Ibrahim Kajo, Yassine Ruichek

Abstract: Highlights•Proposal of a deep loss for image super resolution, denoising, JPEG artifacts removal.•Extend the generative adversarial learning with a degradation estimation CNN network.•Extensive ablation study on super resolution, denoising, JPEG artifacts applications.•The DLL-GAN ensures good performance and robustness on high degradation levels.