Blind image super-resolution with spatially variant degradations

By Victor Cornillère, Abdelaziz Djelouah, Yifan Wang, Olga Sorkine-Hornung, Christopher Schroers. 2019

97 citations recorded by the source corpus.

Existing deep learning approaches to single image super-resolution have achieved impressive results but mostly assume a setting with fixed pairs of high resolution and low resolution images. However, to robustly address realistic upscaling scenarios where the relation between high resolution and low resolution images is unknown, blind image super-resolution is required. To this end, we propose a solution that relies on three components: First, we use a degradation aware SR network to synthesize the HR image given a low resolution image and the corresponding blur kernel. Second, we train a kernel discriminator to analyze the generated high resolution image in order to predict errors present due to providing an incorrect blur kernel to the generator. Finally, we present an optimization procedure that is able to recover both the degradation kernel and the high resolution image by minimizing the error predicted by our kernel discriminator. We also show how to extend our approach to spatially variant degradations that typically arise in visual effects pipelines when compositing content from different sources and how to enable both local and global user interaction in the upscaling process.

Subfields: Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Radu Timofte
2. Luc Van Gool
3. Kai Zhang
4. Jingyun Liang
5. Chen Change Loy
6. Lei Zhang
7. Ming-Hsuan Yang
8. Wangmeng Zuo

In Generative Models

1. Luc Van Gool
2. Ce Zhu
3. Kai Zhang
4. Chao Ren
5. Jingyun Liang
6. Radu Timofte
7. Yuanyuan Wu
8. Xiaohai He

Citations in the corpus

• Designing a Practical Degradation Model for Deep Blind Image Super-Resolution (2021, methodology): “5 , 8] , respectively.”
• Real-World Single Image Super-Resolution: A Brief Review (2021, methodology): “Unlike previous approaches [90], [91] that combine two or more networks, Huang et al. [92] develop a deep alternating network (DAN) for RSISR, in which the iterative optimization process between the super-resolved image and the blur kernel is unfolded to an end-to-end trainable network.”