Research map / Image generation / Editing and style transfer
StatLUT: Statistical Feature-Driven Multimodal 3D LUT Generation for Photorealistic Style Transfer
Yifan Wang, Zhixiang Hao, Yu Wang, Congchao Zhu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes StatLUT, a method that generates 3D lookup tables for photorealistic style transfer from statistical features of multimodal inputs.
Problem
Photorealistic style transfer must match a reference's color and tone while preserving content structure; existing deep methods suffer from semantic entanglement from pre-trained encoders causing spatial distortions, and pixel-level mappings ignore color gamut topology.
Approach
Generates 3D LUTs driven by statistical features rather than semantic encoder features, avoiding semantic entanglement and enforcing color gamut structure; the abstract is truncated so further details are not stated.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace RGB 1D histograms with Lab L, ab and ab-conditioned mean luminance features (ablation) | StatLUT · Feature ablation | Dist. to Ideal: 0.286 -> 0.281 | Improved |
| Add ab-conditioned mean luminance to Lab L and ab histograms (ablation) | StatLUT · Feature ablation | Dist. to Ideal: 0.304 -> 0.281 | Improved |
| Use Lab Lab histograms without abL vs RGB histograms (ablation) | StatLUT · Feature ablation | Style Sim.: 0.874 -> 0.848 | Hurt |
| Use MR-Mapper instead of simple MLP (ablation) | StatLUT · Mapper ablation | Dist. to Ideal: 0.286 -> 0.259 | Improved |
| Use MR-Mapper instead of simple MLP (ablation) | StatLUT · Mapper ablation | parameters (M): 428.97 -> 0.38 | Improved |
| Use MR-Mapper instead of bottleneck MLP (ablation) | StatLUT · Mapper ablation | Content Sim.: 0.726 -> 0.760 | Improved |
| Use MR-Mapper instead of bottleneck MLP (ablation) | StatLUT · Mapper ablation | MACs (M): 5.24 -> 13.82 | Hurt |
| Increase LUT size from 16 to 20 or 24 (ablation) | StatLUT · LUT size ablation | Dist. to Ideal: 0.259 -> 0.268 / 0.260 | No clear effect |
| Increase LUT size from 8 to 16 (ablation) | StatLUT · LUT size ablation | Dist. to Ideal: 0.322 -> 0.259 | Improved |
| Use text-driven H-Diffuser features instead of reference image | StatLUT (Text) · NAS | Style Sim.: 0.746 -> 0.498 | Hurt |
| Use StatLUT image-driven vs NLUT | StatLUT · NAS | Content Sim.: 0.640 -> 0.707 | Improved |
| Use StatLUT vs Neural Preset on content similarity | StatLUT · NAS | Content Sim.: 0.709 -> 0.707 | Hurt |
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.
- SA-LUT — SA-LUT: Spatial Adaptive 4D Look-Up Table for Photorealistic Style Transfer (ICCV 2025) · 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
- SA-LUT: Spatial Adaptive 4D Look-Up Table for Photorealistic Style Transfer (ICCV 2025 · similarity 0.90)
- GLUT: 3D Gaussian Lookup Table for Continuous Color Transformation (NeurIPS 2026 · similarity 0.88)
- CLIPGaussian: Universal and Multimodal Style Transfer Based on Gaussian Splatting (NeurIPS 2025 · similarity 0.87)
- FPRF: Feed-Forward Photorealistic Style Transfer of Large-Scale 3D Neural Radiance Fields (AAAI 2024 · similarity 0.86)
- SGSST: Scaling Gaussian Splatting Style Transfer (CVPR 2025 · similarity 0.85)
- Free-Lunch Color-Texture Disentanglement for Stylized Image Generation (NeurIPS 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.
