mlsanityData to 7 Oct 2026

The Best of Both Worlds: a Framework for Combining Degradation Prediction with High Performance Super-Resolution Networks

arXiv 2211.05018 ↗First posted 9 Nov 2022cs.CV cs.AI cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

To date, the best-performing blind super-resolution (SR) techniques follow one of two paradigms: A) generate and train a standard SR network on synthetic low-resolution - high-resolution (LR - HR) pairs or B) attempt to predict the degradations an LR image has suffered and use these to inform a customised SR network. Despite significant progress, subscribers to the former miss out on useful degradation information that could be used to improve the SR process. On the other hand, followers of the latter rely on weaker SR networks, which are significantly outperformed by the latest architectural advancements. In this work, we present a framework for combining any blind SR prediction mechanism with any deep SR network, using a metadata insertion block to insert prediction vectors into SR network feature maps. Through comprehensive testing, we prove that state-of-the-art contrastive and iterative prediction schemes can be successfully combined with high-performance SR networks such as RCAN and HAN within our framework. We show that our hybrid models consistently achieve stronger SR performance than both their non-blind and blind counterparts. Furthermore, we demonstrate our framework's robustness by predicting degradations and super-resolving images from a complex pipeline of blurring, noise and compression.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 61 of its 98 references that are in the map. First posted on arXiv 9 Nov 2022.

Builds on

61 of 98 references.

References · 61

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  4. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,082
  5. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  6. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  7. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  8. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,482
  9. 2016SGDR: Stochastic Gradient Descent with Warm Restartscited by 4,993
  10. 2017Squeeze-and-Excitation Networkscited by 4,763
  11. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  12. 2020Supervised Contrastive Learningcited by 2,803
38 more
  1. 2020Improved Baselines with Momentum Contrastive Learningcited by 2,282
  2. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,898
  3. 2021SwinIR: Image Restoration Using Swin Transformercited by 1,869
  4. 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,543
  5. 2018Image Super-Resolution Using Very Deep Residual Channel Attention Networkscited by 1,256
  6. 2018ESRGAN: Enhanced Super-Resolution Generative Adversarial Networkscited by 1,250
  7. 2018Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labelscited by 994
  8. 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 987
  9. 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 961
  10. 2021Designing a Practical Degradation Model for Deep Blind Image Super-Resolutioncited by 524
  11. 2022Activating More Pixels in Image Super-Resolution Transformercited by 433
  12. 2019Deep Learning for Image Super-resolution: A Surveycited by 368
  13. 2015Sketch-based Manga Retrieval using Manga109 Datasetcited by 326
  14. 2016Large-Margin Softmax Loss for Convolutional Neural Networkscited by 296
  15. 2017"Zero-Shot" Super-Resolution using Deep Internal Learningcited by 285
  16. 2020Single Image Super-Resolution via a Holistic Attention Networkcited by 265
  17. 2017Learning a Single Convolutional Super-Resolution Network for Multiple Degradationscited by 227
  18. 2019Blind Super-Resolution With Iterative Kernel Correctioncited by 203
  19. 2021Unsupervised Degradation Representation Learning for Blind Super-Resolutioncited by 177
  20. 2019Blind Super-Resolution Kernel Estimation using an Internal-GANcited by 175
  21. 2020CrossTransformers: spatially-aware few-shot transfercited by 174
  22. 2017FSRNet: End-to-End Learning Face Super-Resolution with Facial Priorscited by 170
  23. 2022Efficient Long-Range Attention Network for Image Super-resolutioncited by 151
  24. 2021Transformer for Single Image Super-Resolutioncited by 150
  25. 2018Unsupervised Image Super-Resolution using Cycle-in-Cycle Generative Adversarial Networkscited by 133
  26. 2020Unfolding the Alternating Optimization for Blind Super Resolutioncited by 129
  27. 2019Frequency Separation for Real-World Super-Resolutioncited by 115
  28. 2018Large Margin Deep Networks for Classificationcited by 110
  29. 2018To learn image super-resolution, use a GAN to learn how to do image degradation firstcited by 94
  30. 2021Blind Image Super-Resolution: A Survey and Beyondcited by 93
  31. 2020Unpaired Image Super-Resolution using Pseudo-Supervisioncited by 67
  32. 2021Real-World Single Image Super-Resolution: A Brief Reviewcited by 65
  33. 2021On Efficient Transformer-Based Image Pre-training for Low-Level Visioncited by 64
  34. 2020Unified Dynamic Convolutional Network for Super-Resolution with Variational Degradationscited by 61
  35. 2017Super-FAN: Integrated facial landmark localization and super-resolution of real-world low resolution faces in arbitrary poses with GANscited by 55
  36. 2020Video Super Resolution Based on Deep Learning: A Comprehensive Surveycited by 54
  37. 2022Deep Constrained Least Squares for Blind Image Super-Resolutioncited by 47
  38. 2020KOALAnet: Blind Super-Resolution using Kernel-Oriented Adaptive Local Adjustmentcited by 37

50 most cited of 61 All 61 (JSON)

Built on by

Not cited yet.