Research map / Image generation / Editing and style transfer
Recolour What Matters: Region-Aware Colour Editing via Token-Level Diffusion
Yuqi Yang, Dongliang Chang, Yijia Ling, Ruoyi Du, Zhanyu Ma
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 Lab-space loss (lambda_Lab 0 to 0.8) (ablation) | ColourCrafter · Lab loss total weight ablation | Delta E00: 32.87 -> 28.53 | Improved |
| add Lab-space loss (lambda_Lab 0 to 0.8) (ablation) | ColourCrafter · Lab loss total weight ablation | MAE Hue: 20.48 -> 14.88 | Improved |
| raise lambda_Lab from 0.8 to 1 (ablation) | ColourCrafter · Lab loss total weight ablation | Delta E00: 28.53 -> 37.50 | Hurt |
| add luminance term to Lab loss (lambda_L 0 to 0.5) (ablation) | ColourCrafter · Lab loss luminance weight ablation | Delta E00: 45.15 -> 34.54 | Improved |
| add luminance term to Lab loss (lambda_L 0 to 0.5) (ablation) | ColourCrafter · Lab loss luminance weight ablation | MAE Hue: 34.47 -> 18.49 | Improved |
| raise lambda_L from 0.5 to 0.8 (ablation) | ColourCrafter · Lab loss luminance weight ablation | Delta E00: 34.54 -> 40.79 | Hurt |
| apply Lab loss only inside edit mask (ablation) | ColourCrafter · colour editing | saturation (qualitative) | Improved |
| use RGB loss instead of Lab loss (ablation) | ColourCrafter · colour editing | colour saturation (qualitative) | Hurt |
| inject colour as text tokens instead of fused image tokens (ablation) | ColourCrafter · colour editing | colour stability (qualitative) | Hurt |
| token-level colour fusion versus Flux.1-Kontext text colour name | ColourCrafter · Unsplash colour-editing benchmark | Delta E00: 41.36 -> 28.52 | Improved |
| token-level colour fusion versus Control-Color | ColourCrafter · Unsplash colour-editing benchmark | MAE Hue: 20.31 -> 14.88 | Improved |
| ColourCrafter versus UniEdit-Flow on realism | ColourCrafter · Unsplash colour-editing benchmark | FID: 46.57 vs 51.91 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FlowEdit — FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models (ICCV 2025) · better on 2 of 2 shared results
- ColorPeel — ColorPeel: Color Prompt Learning with Diffusion Models via Color and Shape Disentanglement (ECCV 2024) · better on 2 of 2 shared results
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
- Training-Free Text-Guided Color Editing with Multi-Modal Diffusion Transformer (ICLR 2026 · similarity 0.90)
- NumColor: Precise Numeric Color Control in Text-to-Image Generation (ECCV 2026 · similarity 0.90)
- Automatic Controllable Colorization via Imagination (CVPR 2024 · similarity 0.89)
- ChromaFusionNet (CFNet): Natural Fusion of Fine-Grained Color Editing (AAAI 2024 · similarity 0.88)
- Color Alignment in Diffusion (CVPR 2025 · similarity 0.87)
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 2025 · similarity 0.87)
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.
