Research map / Image generation / Image restoration
High-Resolution Image Harmonization via Collaborative Dual Transformations
Wenyan Cong, Xinhao Tao, Li Niu, Jing Liang, Xuesong Gao, Qihao Sun, Liqing Zhang
Check 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 |
|---|---|---|---|
| Combine low-resolution pixel-to-pixel transformation, high-resolution RGB-to-RGB transformation, and refinement | CDTNet · HAdobe5k 1024×1024 | PSNR: 37.65 -> 38.77 | Improved |
| Add RGB-to-RGB transformation to the pixel-only system (ablation) | CDTNet · HAdobe5k 1024×1024 | PSNR: 29.41 -> 38.77 | Improved |
| Add pixel-to-pixel transformation and refinement to RGB-only transformation (ablation) | CDTNet · HAdobe5k 1024×1024 | PSNR: 37.65 -> 38.77 | Improved |
| Share encoder E between pixel harmonization and RGB weight prediction (ablation) | CDTNet color mapping module · HAdobe5k 1024×1024 | PSNR: 36.86 -> 37.65 | Improved |
| Pool foreground and background features separately before predicting LUT coefficients (ablation) | CDTNet color mapping module · HAdobe5k 1024×1024 | PSNR: 37.31 -> 37.65 | Improved |
| Include the high-resolution RGB transformation result in refinement (ablation) | CDTNet refinement module · HAdobe5k 1024×1024 | PSNR: 37.29 -> 38.77 | Improved |
| Include the upsampled pixel-to-pixel result in refinement (ablation) | CDTNet refinement module · HAdobe5k 1024×1024 | PSNR: 37.78 -> 38.77 | Improved |
| Include the high-resolution mask in refinement (ablation) | CDTNet refinement module · HAdobe5k 1024×1024 | PSNR: 37.33 -> 38.77 | Improved |
| Include the upsampled decoder feature in refinement (ablation) | CDTNet refinement module · HAdobe5k 1024×1024 | PSNR: 37.64 -> 38.77 | Improved |
| Use a blending layer in refinement (ablation) | CDTNet refinement module · HAdobe5k 1024×1024 | PSNR: 28.13 -> 38.77 | Improved |
| Use deep RGB-to-RGB transformation only at inference instead of the full collaborative method (ablation) | CDTNet-256 · HAdobe5k 1024×1024 | PSNR: 38.77 -> 37.65 | Hurt |
| Use deep RGB-to-RGB transformation only at inference instead of the full collaborative method (ablation) | CDTNet-256 · HAdobe5k 2048×2048 | PSNR: 37.66 -> 37.28 | Hurt |
Closest papers by meaning
- High-Resolution Image Harmonization with Adaptive-Interval Color Transformation (NeurIPS 2024 · similarity 0.88)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.88)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · similarity 0.86)
- Prior-guided Hierarchical Harmonization Network for Efficient Image Dehazing (AAAI 2025 · similarity 0.84)
- Training-and-Prompt-Free General Painterly Harmonization via Zero-Shot Disentenglement on Style and Content References (AAAI 2025 · similarity 0.84)
- Relightful Harmonization: Lighting-aware Portrait Background Replacement (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.
