Pairwise Distance Distillation for Unsupervised Real-World Image Super-Resolution
This addresses the problem of real-world image super-resolution for applications like photography or surveillance, where paired training data is unavailable and degradations are unknown, though it is incremental as it builds on existing distillation and unsupervised techniques.
The paper tackles unsupervised real-world image super-resolution with unknown degradations by introducing a pairwise distance distillation framework that adapts a synthetically-trained model to target real-world degradations, significantly enhancing fidelity and perceptual quality and surpassing state-of-the-art methods.
Standard single-image super-resolution creates paired training data from high-resolution images through fixed downsampling kernels. However, real-world super-resolution (RWSR) faces unknown degradations in the low-resolution inputs, all the while lacking paired training data. Existing methods approach this problem by learning blind general models through complex synthetic augmentations on training inputs; they sacrifice the performance on specific degradation for broader generalization to many possible ones. We address the unsupervised RWSR for a targeted real-world degradation. We study from a distillation perspective and introduce a novel pairwise distance distillation framework. Through our framework, a model specialized in synthetic degradation adapts to target real-world degradations by distilling intra- and inter-model distances across the specialized model and an auxiliary generalized model. Experiments on diverse datasets demonstrate that our method significantly enhances fidelity and perceptual quality, surpassing state-of-the-art approaches in RWSR. The source code is available at https://github.com/Yuehan717/PDD.