Research map / Image generation / Image restoration
Harmonizer: Learning to Perform White-Box Image and Video Harmonization
Zhanghan Ke, Chunyi Sun, Lei Zhu, Ke Xu, Rynson W.H. Lau
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add cascade regressor (ablation) | Harmonizer · iHarmony4 All (256x256) | MSE: 28.47 -> 26.85 | Improved |
| add dynamic loss strategy on top of cascade regressor (ablation) | Harmonizer · iHarmony4 All (256x256) | MSE: 26.85 -> 24.26 | Improved |
| add both cascade regressor and dynamic loss (ablation) | Harmonizer · iHarmony4 All (256x256) | PSNR: 36.96 -> 37.84 | Improved |
| increase number of filters from 2 to 4 (ablation) | Harmonizer · iHarmony4 All (256x256) | MSE: 74.16 -> 29.60 | Improved |
| increase number of filters from 4 to 6 (ablation) | Harmonizer · iHarmony4 All (256x256) | MSE: 29.60 -> 24.26 | Improved |
| increase number of filters from 6 to 8 (ablation) | Harmonizer · iHarmony4 All (256x256) | MSE: 24.26 -> 23.51 | Mixed |
| replace autoencoder with white-box filter regression (without cascade or dynamic loss) | Harmonizer vs IHT · iHarmony4 All (256x256) | MSE: 37.07 -> 33.28 | Improved |
| evaluate at original resolution instead of 256x256 | IHT · iHarmony4 All | MSE: 37.07 -> 57.22 | Hurt |
| evaluate at original resolution instead of 256x256 | Harmonizer · iHarmony4 All | MSE: 24.26 -> 27.62 | Mixed |
| add Polynomial Color Mapping upsampling to DOVE | DOVE · iHarmony4 All (original resolution) | MSE: 72.98 -> 67.35 | Mixed |
| add Polynomial Color Mapping upsampling to IHT | IHT · iHarmony4 All (original resolution) | MSE: 58.89 -> 57.22 | Mixed |
| use Harmonizer instead of IHT for video harmonization user study | Harmonizer · User study B-T score (Video) | B-T score: 0.601 -> 2.042 | Improved |
Later papers that compared with this method
6 later accepted papers report Harmonizer as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- LBM: Latent Bridge Matching for Fast Image-to-Image Translation (ICCV 2025 · better on 1 of 1 shared result)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · better on 6 of 10 shared results)
- Relightful Harmonization: Lighting-aware Portrait Background Replacement (CVPR 2024 · better on 4 of 4 shared results)
- Semi-Supervised Parametric Real-World Image Harmonization (CVPR 2023 · better on 7 of 8 shared results)
- Learning Global-aware Kernel for Image Harmonization (ICCV 2023 · better on 0 of 2 shared results)
- Deep Image Harmonization with Globally Guided Feature Transformation and Relation Distillation (ICCV 2023 · better on 0 of 1 shared result)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Transformer for Image Harmonization and Beyond (TPAMI 2022 · similarity 0.88)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.87)
- Training-and-Prompt-Free General Painterly Harmonization via Zero-Shot Disentenglement on Style and Content References (AAAI 2025 · similarity 0.86)
- Prior-guided Hierarchical Harmonization Network for Efficient Image Dehazing (AAAI 2025 · similarity 0.86)
- Painterly Image Harmonization by Learning from Painterly Objects (AAAI 2024 · similarity 0.85)
- Video Harmonization with Triplet Spatio-Temporal Variation Patterns (CVPR 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
