mlsanityData to 7 Oct 2026

Normalizing Flow as a Flexible Fidelity Objective for Photo-Realistic Super-resolution

arXiv 2111.03649 ↗First posted 5 Nov 2021cs.CV eess.IVcited by 10

Not on the map. Why?

Abstract (arXiv)

Super-resolution is an ill-posed problem, where a ground-truth high-resolution image represents only one possibility in the space of plausible solutions. Yet, the dominant paradigm is to employ pixel-wise losses, such as L_1, which drive the prediction towards a blurry average. This leads to fundamentally conflicting objectives when combined with adversarial losses, which degrades the final quality. We address this issue by revisiting the L_1 loss and show that it corresponds to a one-layer conditional flow. Inspired by this relation, we explore general flows as a fidelity-based alternative to the L_1 objective. We demonstrate that the flexibility of deeper flows leads to better visual quality and consistency when combined with adversarial losses. We conduct extensive user studies for three datasets and scale factors, where our approach is shown to outperform state-of-the-art methods for photo-realistic super-resolution. Code and trained models will be available at: git.io/AdFlow

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 33 of its 49 references that are in the map. First posted on arXiv 5 Nov 2021.

Builds on

33 of 49 references.

References · 33

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  3. 2016Image-to-Image Translation with Conditional Adversarial Networkscited by 4,510
  4. 2015Variational Inference with Normalizing Flowscited by 2,161
  5. 2016Density estimation using Real NVPcited by 2,029
  6. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,898
  7. 2017Enhanced Deep Residual Networks for Single Image Super-Resolutioncited by 1,543
  8. 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,539
  9. 2018Image Super-Resolution Using Very Deep Residual Channel Attention Networkscited by 1,256
  10. 2018ESRGAN: Enhanced Super-Resolution Generative Adversarial Networkscited by 1,250
  11. 2014Image Super-Resolution Using Deep Convolutional Networkscited by 1,218
  12. 2014NICE: Non-linear Independent Components Estimationcited by 1,052
21 more
  1. 2015Accurate Image Super-Resolution Using Very Deep Convolutional Networkscited by 902
  2. 2017The Perception-Distortion Tradeoffcited by 582
  3. 2017Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolutioncited by 452
  4. 2015Deep multi-scale video prediction beyond mean square errorcited by 297
  5. 2018Deep Back-Projection Networks For Super-Resolutioncited by 224
  6. 2020SRFlow: Learning the Super-Resolution Space with Normalizing Flowcited by 209
  7. 2016EnhanceNet: Single Image Super-Resolution Through Automated Texture Synthesiscited by 177
  8. 2019Learning Likelihoods with Conditional Normalizing Flowscited by 170
  9. 2018The 2018 PIRM Challenge on Perceptual Image Super-resolutioncited by 155
  10. 2019Guided Image Generation with Conditional Invertible Neural Networkscited by 147
  11. 2019RankSRGAN: Generative Adversarial Networks with Ranker for Image Super-Resolutioncited by 145
  12. 2019Frequency Separation for Real-World Super-Resolutioncited by 115
  13. 2020Invertible Image Rescalingcited by 92
  14. 2020NTIRE 2020 Challenge on Real-World Image Super-Resolution: Methods and Resultscited by 83
  15. 2019Progressive Face Super-Resolution via Attention to Facial Landmarkcited by 73
  16. 2021Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescalingcited by 73
  17. 2019Unsupervised Learning for Real-World Super-Resolutioncited by 64
  18. 2019C-Flow: Conditional Generative Flow Models for Images and 3D Point Cloudscited by 50
  19. 2019Explorable Super Resolutioncited by 44
  20. 2019AIM 2019 Challenge on Real-World Image Super-Resolution: Methods and Resultscited by 34
  21. 2021DeFlow: Learning Complex Image Degradations from Unpaired Data with Conditional Flowscited by 30

Built on by

Cited by year

Citing papers per year: 2022: 3; 2023: 3; 2024: 3; 2025: 1.

10 papers.

  1. 2023Local Implicit Normalizing Flow for Arbitrary-Scale Image Super-Resolutioncited by 34
  2. 2022Collapse by Conditioning: Training Class-conditional GANs with Limited Datacited by 24
  3. 2025Reversing Flow for Image Restorationcited by 10
  4. 2022On the Robustness of Normalizing Flows for Inverse Problems in Imagingcited by 8
  5. 2024Boosting Flow-based Generative Super-Resolution Models via Learned Priorcited by 7
  6. 2022Invertible Monotone Operators for Normalizing Flowscited by 6
  7. 2023Learning Many-to-Many Mapping for Unpaired Real-World Image Super-resolution and Downscalingcited by 6
  8. 2024Self-Adaptive Reality-Guided Diffusion for Artifact-Free Super-Resolutioncited by 6
  9. 2023Towards Climate Variable Prediction with Conditioned Spatio-Temporal Normalizing Flowscited by 0
  10. 2024A General Method to Incorporate Spatial Information into Loss Functions for GAN-based Super-resolution Modelscited by 0