Deep Blind Un-Supervised Learning Network For Single Image Super Resolution

Recently, image super-resolution (SR), which predicts high-resolution (HR) images from low-resolution (LR) images, has shown increasing performance improvements with the progress of deep learning. Most of the current methods have struggled to design more complicated and deeper network architectures and aimed to learn a good LR-to-HR mapping with the previously prepared training sample pairs under a fixed degradation model (Blurring and down-sampling operations) such as bicubic dawn-sampling. However, these methods cannot be generalized to real scenarios with unknown complex degradation models. In this study, we propose a blind unsupervised learning network for automatically estimating the unknown degradation operations in a single SR problem. Motivated by the considerable possessed image priors in the network architectures, we construct a generative network for simultaneously learning the inherent priors of the latent LR low-resolution observation only. Experimental results on two benchmark datasets demonstrate that the proposed method shows promising performance under unknown degradation models.