Research map / 3D vision / Gaussian splatting
3D Skew Normal Splatting
Read the paperCheck your idea against 108,967 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 Skew-Normal Splatting (SNS), replacing the Gaussian primitive in 3D Gaussian Splatting with a bounded, learnable-skewness Azzalini skew-normal kernel with closed-form 2D projection and block-wise optimization.
Problem
Symmetric Gaussian primitives need many overlapping kernels to represent sharp boundaries and one-sided structures under a finite primitive budget; prior non-Gaussian kernels (GES, SSS) remain elliptical and HGS uses hard truncation.
Approach
Use the Azzalini skew-normal distribution as the splatting kernel with a bounded skew vector k, decoupled from rotation/scale via an reparameterization, closed under affine projection so it fits standard tile-based rasterization, plus a corrected bounding-box center and block-coordinate alternating optimization of shape vs. skew parameters.
Key results
- On Mip-NeRF 360 SNS reaches 30.17 PSNR vs 29.90 for SSS (second best) and 28.69 for 3DGS.
- On Tanks&Temples SNS reaches 25.08 PSNR, 0.877 SSIM, 0.132 LPIPS, improving over SSS (24.87, 0.873, 0.138).
- On Deep Blending SNS reaches 30.30 PSNR, best among compared methods; ablation shows alternating optimization and k-decomposition give moderate gains.
Surprising findings
- Individual ablation variants sometimes outperform the full SNS model on SSIM and LPIPS, so the full model is not uniformly best.
Limitations
- Evaluating the skew-normal CDF via erf incurs a moderate rendering speed penalty.
- Does not address highly specular surfaces since SH color modeling is retained.
- Full model does not uniformly beat ablated variants on SSIM and LPIPS.
Experimental setup
Datasets: Mip-NeRF 360 Tanks&Temples Deep Blending
Benchmarks: Mip-NeRF 360 Tanks&Temples Deep Blending
Metrics: PSNR SSIM LPIPS
Baselines: Mip-NeRF 3DGS GES 3D-HGS Fre-GS Scaffold-GS 3DGS-MCMC SSS
Builds on: 3D Gaussian Splatting Skew-Normal distribution (Azzalini) Student Splatting and Scooping Generalized Exponential Splatting 3D-HGS
Closest papers by meaning
- Effective Rank Analysis and Regularization for Enhanced 3D Gaussian Splatting (NeurIPS 2024 · similarity 0.91)
- Normal-GS: 3D Gaussian Splatting with Normal-Involved Rendering (NeurIPS 2024 · similarity 0.90)
- ReorgGS: Equivalent Distribution Reorganization for 3D Gaussian Splatting (NeurIPS 2026 · similarity 0.90)
- IBGS: Image-Based Gaussian Splatting (NeurIPS 2025 · similarity 0.90)
- GaussianSpa: An "Optimizing-Sparsifying" Simplification Framework for Compact and High-Quality 3D Gaussian Splatting (CVPR 2025 · similarity 0.90)
- GaussianPro: 3D Gaussian Splatting with Progressive Propagation (ICML 2024 · similarity 0.90)
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.