Research map / 3D vision / Gaussian splatting
NURBS Splatting: A Unified Differentiable Rendering Framework for Vector Graphics
Jingye Qiu, Shizhe Zhou
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 |
|---|---|---|---|
| replace B-spline/DiffVG baseline with NURBS Splatting (weights, knots, Gaussian rendering) | calligraphy reconstruction · Calligraphy reconstruction (191 chars) | PSNR: 24.72 -> 26.32 | Improved |
| use NURBS Splatting instead of Berio et al. on SSIM | calligraphy reconstruction · Calligraphy reconstruction (191 chars) | SSIM: 0.9813 -> 0.9793 | Hurt |
| increase contour density from 18 to 30 (ablation) | NURBS Splatting · Calligraphy reconstruction (191 chars) | SSIM: 0.9793 -> 0.9824 | Improved |
| add learnable rational weights (ablation) | NURBS Splatting · Calligraphy reconstruction (191 chars) | PSNR: 25.63 -> 26.32 | Improved |
| add learnable non-uniform knots (ablation) | NURBS Splatting · Calligraphy reconstruction (191 chars) | PSNR: 26.07 -> 26.32 | Improved |
| add learnable knots (cost in time) (ablation) | NURBS Splatting · Calligraphy reconstruction (191 chars) | optimization time (s): 5.63 -> 8.73 | Hurt |
| Gaussian splatting backend with uniform B-splines vs DiffVG baseline (ablation) | Berio et al. B-spline baseline · Calligraphy reconstruction (191 chars) | optimization time (s): 10.65 -> 5.12 | Improved |
| use isotropic Gaussians instead of tangent-aligned anisotropic Gaussians (ablation) | NURBS Splatting (no weights/knots) · Calligraphy reconstruction (191 chars) | PSNR: 18.75 -> 25.92 | Improved |
| replace LIVE DiffVG Bezier paths with NURBS Splatting | LIVE · Layer-wise vectorization (100 Noto Emoji) | MSE: 0.0025 -> 0.0017 | Improved |
| add grid-step annealing for filled regions (ablation) | NURBS Splatting (LIVE) · Layer-wise vectorization (100 Noto Emoji) | time (s): 527 -> 410 s, MSE 0.0015 -> 0.0017 | Mixed |
| add rational weights and knots (ablation) | NURBS Splatting (LIVE) · Layer-wise vectorization (100 Noto Emoji) | MSE: 0.0018 -> 0.0017 | Improved |
| NURBS Splatting vs Berio et al. in rendering speed at low resolution | efficiency test · Runtime, closed curves 512x512, 16 curves | total time (ms): 100.99 -> 135.24 | Hurt |
Closest papers by meaning
- Bézier Splatting for Fast and Differentiable Vector Graphics Rendering (NeurIPS 2025 · similarity 0.89)
- DisC-GS: Discontinuity-aware Gaussian Splatting (NeurIPS 2024 · similarity 0.87)
- Sort-free Gaussian Splatting via Weighted Sum Rendering (ICLR 2025 · similarity 0.86)
- Anti-Aliased 2D Gaussian Splatting (NeurIPS 2025 · similarity 0.86)
- Evolving High-Quality Rendering and Reconstruction in a Unified Framework with Contribution-Adaptive Regularization (CVPR 2025 · similarity 0.86)
- UniGS: Unified Geometry-Aware Gaussian Splatting for Multimodal Rendering (NeurIPS 2026 · similarity 0.86)
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.
