Research map / Image generation / Image restoration
CAST-LUT: Tokenizer-Guided HSV Look-Up Tables for Purple Flare Removal
Pu Wang, Shuning Sun, Jialang Lu, Chen Wu, Zhihua Zhang, Youshan Zhang, Chenggang Shan, Dianjie Lu, Guijuan Zhang, Zhuoran Zheng
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 decoupled HSV 1D-LUTs instead of RGB 1D-LUTs (ablation) | CAST-LUT · PFSD test | PSNR-F: 24.15 -> 30.74 | Improved |
| use decoupled 1D-LUTs instead of 3D RGB-LUT (ablation) | CAST-LUT · PFSD test | HAE: 7.79 -> 4.1 | Improved |
| use CAST tokenizer instead of a simple CNN encoder (ablation) | CAST-LUT · PFSD test | PSNR-F: 29.01 -> 30.74 | Improved |
| add vector quantization to CAST tokens (ablation) | CAST-LUT · PFSD test | PSNR-F: 28.45 -> 30.74 | Improved |
| add residual branch (ablation) | CAST-LUT · PFSD test | PSNR-NF: 27.37 -> 34.35 | Improved |
| use NL=16 fused LUT groups instead of 1 (ablation) | CAST-LUT · PFSD test | PSNR: 30.11 -> 34.96 | Improved |
| increase NL from 16 to 32 (ablation) | CAST-LUT · PFSD test | PSNR: 34.96 -> 32.43 | Hurt |
| add flare suppression loss Lf (ablation) | CAST-LUT · PFSD test | HAE: 7.2 -> 4.1 | Improved |
| add perceptual loss Lp (ablation) | CAST-LUT · PFSD test | LPIPS: 0.061 -> 0.03 | Improved |
| add perceptual loss Lp (ablation) | CAST-LUT · PFSD test | HAE: 4.2 -> 4.1 | No clear effect |
| use CAST-LUT instead of NILUT | LUT-based flare removal · PFSD test | HAE: 4.87 -> 4.10 | Improved |
| use CAST-LUT instead of BPAM in compute cost | flare removal · PFSD test | FLOPs (G): 15.21 -> 23.32 | Hurt |
Closest papers by meaning
- DCA-LUT: Deep Chromatic Alignment with 5D LUT for Purple Fringing Removal (AAAI 2026 · similarity 0.89)
- Light Source Guided Single-Image Flare Removal From Unpaired Data (ICCV 2021 · similarity 0.87)
- Luminance-Aware Statistical Quantization: Unsupervised Hierarchical Learning for Illumination Enhancement (NeurIPS 2025 · similarity 0.85)
- GLUT: 3D Gaussian Lookup Table for Continuous Color Transformation (NeurIPS 2026 · similarity 0.85)
- Bright-NeRF: Brightening Neural Radiance Field with Color Restoration from Low-Light RAW Images (AAAI 2025 · similarity 0.85)
- Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustment (CVPR 2025 · 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.
