Research map / Learning theory and algorithms / Matrix and tensor methods
Semi-Global Weighted Least Squares in Image Filtering
Wei Liu, Xiaogang Chen, Chuanhua Shen, Zhi Liu, Jie Yang
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 |
|---|---|---|---|
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Art | MAD: 2.4 -> 1.9 | Improved |
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Book | MAD: 1.31 -> 1.12 | Improved |
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Dolls | MAD: 1.37 -> 1.11 | Improved |
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Laundry | MAD: 1.6 -> 1.17 | Improved |
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Moebius | MAD: 1.37 -> 1.08 | Improved |
| enlarge neighborhood to r=4 and stride tau=4 (from r=1, tau=1) (ablation) | SG-WLS · Guided depth upsampling 4x, Reindeer | MAD: 1.68 -> 1.32 | Improved |
| use r=4 with smaller lambda instead of r=1 with large lambda (ablation) | SG-WLS · Guided upsampling 4x (Fig. 4) | MAD: 2.4 -> 1.9 | Improved |
| increase sliding step tau up to r (ablation) | SG-WLS · Guided depth upsampling / detail enhancement | MAD | No clear effect |
| increase tau to reduce number of subsystems (ablation) | SG-WLS · Smoothing 1-megapixel RGB image | computation time | Improved |
| use 2D neighborhood subsystem construction instead of FGS 1D neighbors | SG-WLS (r=1, tau=1) · Detail enhancement and HDR tone mapping | visual artifacts | Improved |
| replace WLS by SG-WLS | SG-WLS (r=1, tau=1) · 1-megapixel RGB smoothing | time (seconds): 8.02 -> 0.35 | Improved |
| SG-WLS (r=1,tau=1) versus FGS | SG-WLS · 1-megapixel RGB smoothing | time (seconds): 0.1 -> 0.35 | Hurt |
Closest papers by meaning
- Fast Global Image Smoothing via Quasi Weighted Least Squares (IJCV 2024 · similarity 0.85)
- GFNet: Global Filter Networks for Visual Recognition (TPAMI 2023 · similarity 0.84)
- Reweighted Solutions for Weighted Low Rank Approximation (ICML 2024 · similarity 0.83)
- WSVD: Weighted Low-Rank Approximation for Fast and Efficient Execution of Low-Precision Vision-Language Models (ICLR 2026 · similarity 0.83)
- Precise asymptotics of reweighted least-squares algorithms for linear diagonal networks (NeurIPS 2024 · similarity 0.83)
- Recursive Generalization Transformer for Image Super-Resolution (ICLR 2024 · similarity 0.82)
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.
