Research map / Image generation / Image restoration
Spatial-Separated Curve Rendering Network for Efficient and High-Resolution Image Harmonization
Jingtang Liang, Xiaodong Cun, Chi-Man Pun, Jue Wang
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 |
|---|---|---|---|
| use relative L1 loss on the masked region instead of plain L1 (ablation) | S2CRNet-SqueezeNet · iHarmony4 PSNR | PSNR: 32.11 -> 36.20 | Improved |
| add adversarial loss (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 36.20 -> 36.27 | Improved |
| add adversarial loss (ablation) | S2CRNet-SqueezeNet · HCOCO | PSNR: 37.59 -> 37.51 | No clear effect |
| use spatial-separated encoder instead of foreground-only features (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 34.57 -> 36.27 | Improved |
| use spatial-separated encoder instead of full-image features (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 33.88 -> 36.27 | Improved |
| add cascaded refinement to CRM (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 36.27 -> 36.41 | Improved |
| add semantic labels to cascaded CRM (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 36.41 -> 36.45 | Improved |
| use piece-wise linear curve instead of Zero-DCE curve (ablation) | CRM · iHarmony4 | PSNR: 34.39 -> 36.27 | Improved |
| use piece-wise linear curve instead of 3DLUT (ablation) | CRM · iHarmony4 | PSNR: 35.49 -> 36.27 | Improved |
| set curve levels L=64 instead of 32 (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 36.17 -> 36.45 | Improved |
| increase curve levels L from 64 to 128 (ablation) | S2CRNet-SqueezeNet · iHarmony4 | PSNR: 36.45 -> 36.19 | Hurt |
| use VGG16 backbone instead of SqueezeNet | S2CRNet · iHarmony4 | PSNR: 36.45 -> 37.18 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- DoveNet — DoveNet: Deep Image Harmonization via Domain Verification (CVPR 2020) · 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
- Curve-Aware Gaussian Splatting for 3D Parametric Curve Reconstruction (ICCV 2025 · similarity 0.85)
- Progressive Painterly Image Harmonization from Low-Level Styles to High-Level Styles (AAAI 2024 · similarity 0.85)
- High-Resolution Image Harmonization via Collaborative Dual Transformations (CVPR 2022 · similarity 0.85)
- FRIH: Fine-Grained Region-Aware Image Harmonization (AAAI 2024 · similarity 0.84)
- Feature 3DGS: Supercharging 3D Gaussian Splatting to Enable Distilled Feature Fields (CVPR 2024 · similarity 0.83)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 2024 · similarity 0.83)
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.
