Research map / Image generation / Editing and style transfer
SA-LUT: Spatial Adaptive 4D Look-Up Table for Photorealistic Style Transfer
Zerui Gong, Zhonghua Wu, Qingyi Tao, Qinyue Li, Chen Change Loy
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 Context Generator (4D LUT) instead of standard 3D LUT (ablation) | SA-LUT · PST50 | LPIPS: 0.14 -> 0.12 | Improved |
| add Context Generator (4D LUT) instead of standard 3D LUT (ablation) | SA-LUT · PST50 | H-Corr: 0.38 -> 0.51 | Improved |
| add cross-attention between content and style for context map (ablation) | SA-LUT · PST50 | LPIPS: 0.13 -> 0.12 | Improved |
| add cross-attention between content and style for context map (ablation) | SA-LUT · PST50 | H-Corr: 0.46 -> 0.51 | Improved |
| use 64 basis LUTs instead of 32 (ablation) | SA-LUT · PST50 | H-Corr: 0.41 -> 0.51 | Improved |
| increase basis LUTs from 64 to 128 (ablation) | SA-LUT · PST50 | H-Corr: 0.51 -> 0.47 | Hurt |
| increase basis LUTs from 64 to 256 (ablation) | SA-LUT · PST50 | H-Corr: 0.51 -> 0.39 | Hurt |
| add real-style adversarial stream to synthetic-only training (ablation) | SA-LUT · PST50 | H-Corr: 0.44 -> 0.51 | Improved |
| add synthetic stream to real-style-only training (ablation) | SA-LUT · PST50 | LPIPS | Improved |
| SA-LUT versus NLUT LUT generation | LUT-based style transfer · PST50 | LUT generation time (s): 16.1112 -> 0.2128 | Improved |
| SA-LUT versus NLUT perceptual distance | LUT-based style transfer · PST50 | LPIPS: 0.36 -> 0.12 | Improved |
Later papers that compared with this method
5 later accepted papers report SA-LUT as a baseline in their results tables; 5 of them report a better number on most of the results they share with it.
- AceTone: Bridging Words and Colors for Conditional Image Grading (CVPR 2026 · better on 8 of 8 shared results)
- Hist2Style: Histogram-Guided Stylization with Bilateral Grids (CVPR 2026 · better on 4 of 4 shared results)
- Towards In-Context Tone Style Transfer with A Large-Scale Triplet Dataset (ECCV 2026 · better on 3 of 4 shared results)
- ColorFM: An Optimization-to-Learning Framework for Color Transfer via Flow Matching (ECCV 2026 · better on 2 of 2 shared results)
- StatLUT: Statistical Feature-Driven Multimodal 3D LUT Generation for Photorealistic Style Transfer (NeurIPS 2026 · better on 1 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
- StatLUT: Statistical Feature-Driven Multimodal 3D LUT Generation for Photorealistic Style Transfer (NeurIPS 2026 · similarity 0.90)
- GLUT: 3D Gaussian Lookup Table for Continuous Color Transformation (NeurIPS 2026 · similarity 0.87)
- Learning Dynamic Style Kernels for Artistic Style Transfer (CVPR 2023 · similarity 0.87)
- Ultrafast Photorealistic Style Transfer via Neural Architecture Search (AAAI 2020 · similarity 0.86)
- CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- SGSST: Scaling Gaussian Splatting Style Transfer (CVPR 2025 · similarity 0.85)
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.
