mlsanityData to 7 Oct 2026

Semi-supervised Parametric Real-world Image Harmonization

arXiv 2303.00157 ↗First posted 1 Mar 2023cs.CVcited by 26

Not on the map. Why?

Abstract (arXiv)

Learning-based image harmonization techniques are usually trained to undo synthetic random global transformations applied to a masked foreground in a single ground truth photo. This simulated data does not model many of the important appearance mismatches (illumination, object boundaries, etc.) between foreground and background in real composites, leading to models that do not generalize well and cannot model complex local changes. We propose a new semi-supervised training strategy that addresses this problem and lets us learn complex local appearance harmonization from unpaired real composites, where foreground and background come from different images. Our model is fully parametric. It uses RGB curves to correct the global colors and tone and a shading map to model local variations. Our method outperforms previous work on established benchmarks and real composites, as shown in a user study, and processes high-resolution images interactively.

Key facts. Cited by 26 papers in the map (data through 7 Oct 2026). Builds on 24 of its 40 references that are in the map. First posted on arXiv 1 Mar 2023.

Builds on

24 of 40 references.

References · 24

  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. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  5. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  6. 2022Generative Adversarial Networkscited by 5,603
  7. 2021Real-ESRGAN: Training Real-World Blind Super-Resolution with Pure Synthetic Datacited by 961
  8. 2021Resolution-robust Large Mask Inpainting with Fourier Convolutionscited by 730
  9. 2019Detecting and Simulating Artifacts in GAN Fake Imagescited by 293
  10. 2017Deep Bilateral Learning for Real-Time Image Enhancementcited by 200
  11. 2020Learning Image-adaptive 3D Lookup Tables for High Performance Photo Enhancement in Real-timecited by 163
  12. 2019Single Image Portrait Relightingcited by 152
12 more
  1. 2019DoveNet: Deep Image Harmonization via Domain Verificationcited by 134
  2. 2017Deep Image Harmonizationcited by 120
  3. 2020DeepLPF: Deep Local Parametric Filters for Image Enhancementcited by 107
  4. 2022Learning to Relight Portrait Images via a Virtual Light Stage and Synthetic-to-Real Adaptationcited by 67
  5. 2021Making Images Real Again: A Comprehensive Survey on Deep Image Compositioncited by 63
  6. 2022Harmonizer: Learning to Perform White-Box Image and Video Harmonizationcited by 61
  7. 2021SSH: A Self-Supervised Framework for Image Harmonizationcited by 60
  8. 2021High-Resolution Image Harmonization via Collaborative Dual Transformationscited by 55
  9. 2015Learning a Discriminative Model for the Perception of Realism in Composite Imagescited by 52
  10. 2021Free-viewpoint Indoor Neural Relighting from Multi-view Stereocited by 50
  11. 2022DCCF: Deep Comprehensible Color Filter Learning Framework for High-Resolution Image Harmonizationcited by 46
  12. 2021Spatial-Separated Curve Rendering Network for Efficient and High-Resolution Image Harmonizationcited by 25

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 11; 2024: 4; 2025: 9; 2026: 2.

26 papers.

Sort:

By area:

  1. 2023SmartEdit: Exploring Complex Instruction-based Image Editing with Multimodal Large Language Modelscited by 174
  2. 2023Relightful Harmonization: Lighting-aware Portrait Background Replacementcited by 56
  3. 2023ControlCom: Controllable Image Composition using Diffusion Modelcited by 43
  4. 2025LBM: Latent Bridge Matching for Fast Image-to-Image Translationcited by 36
  5. 2023Dense Pixel-to-Pixel Harmonization via Continuous Image Representationcited by 19
  6. 2023Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistencycited by 19
  7. 2023DreamCom: Finetuning Text-guided Inpainting Model for Image Compositioncited by 17
  8. 2023DiFaReli++: Diffusion Face Relighting with Consistent Cast Shadowscited by 14
  9. 2024DeCo: Decoupled Human-Centered Diffusion Video Editing with Motion Consistencycited by 13
  10. 2025DreamFuse: Adaptive Image Fusion with Diffusion Transformercited by 13
  11. 2025Comprehensive Relighting: Generalizable and Consistent Monocular Human Relighting and Harmonizationcited by 12
  12. 2023Intrinsic Harmonization for Illumination-Aware Compositingcited by 11
14 more
  1. 2023Zero-Shot Image Harmonization with Generative Model Priorcited by 8
  2. 2025MV-CoLight: Efficient Object Compositing with Consistent Lighting and Shadow Generationcited by 8
  3. 2025DreamLight: Towards Harmonious and Consistent Image Relightingcited by 8
  4. 2024Training-Free Semantic Video Composition via Pre-trained Diffusion Modelcited by 4
  5. 2023Painterly Image Harmonization by Learning from Painterly Objectscited by 3
  6. 2025Omni-Dish: Photorealistic and Faithful Image Generation and Editing for Arbitrary Chinese Dishescited by 3
  7. 2025Gen-n-Val: Agentic Image Data Generation and Validationcited by 3
  8. 2023Progressive Painterly Image Harmonization from Low-level Styles to High-level Stylescited by 1
  9. 2024Diverse Image Harmonizationcited by 1
  10. 2025From Gallery to Wrist: Realistic 3D Bracelet Insertion in Videoscited by 1
  11. 2025POLAR: A Portrait OLAT Dataset and Generative Framework for Illumination-Aware Face Modelingcited by 1
  12. 2024Inverse Entropic Optimal Transport Solves Semi-supervised Learning via Data Likelihood Maximizationcited by 0
  13. 2026Aligned Stable Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistencycited by 0
  14. 2026Cinematic Compositing Using Character-Environment-Harmonized Video Generation Modelscited by 0