Research map / 3D vision / Gaussian splatting
Revisiting 3D Reconstruction Kernels as Low-Pass Filters
Shengjun Zhang, Min Chen, Yibo Wei, Mingyu Dong, Yueqi Duan
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 Gaussian kernel with Jinc kernel | 3DGS · NeRF Synthetic 64x64 | PSNR: 24.01 -> 29.87 | Improved |
| replace Student's t kernel with Jinc kernel | SSS · NeRF Synthetic 64x64 | PSNR: 29.17 -> 29.87 | Improved |
| replace Student's t kernel with Jinc kernel | SSS · NeRF Synthetic 128x128 | LPIPS: 0.0230 -> 0.0411 | Hurt |
| replace Gaussian kernel with Jinc kernel | 3DGS · NeRF Synthetic 128x128 | PSNR: 22.30 -> 31.40 | Improved |
| add frequency modulation to Gaussian kernel | 3DGS · Mip-NeRF360 | PSNR: 28.69 -> 29.16 | Improved |
| add frequency modulation to Gaussian kernel | 3DGS · Tanks & Temples | PSNR: 23.14 -> 23.86 | Improved |
| add frequency modulation to Gaussian kernel | 3DGS · Deep Blending | PSNR: 29.41 -> 29.84 | Improved |
| add frequency modulation to Student's t kernel | SSS · Mip-NeRF360 | PSNR: 29.90 -> 29.96 | Improved |
| add frequency modulation to Student's t kernel | SSS · Tanks & Temples | PSNR: 24.87 -> 24.95 | Improved |
| add frequency modulation to Student's t kernel | SSS · Deep Blending | LPIPS: 0.247 -> 0.242 | Improved |
| use truncated Jinc instead of Gaussian kernel when evaluating at 1/8 resolution after full-res training | 3DGS · Tanks & Temples 1/8 res | PSNR: 15.77 -> 16.47 | Improved |
| use truncated Jinc instead of Gaussian kernel at full resolution | 3DGS · Tanks & Temples full res | PSNR: 23.14 -> 22.55 | 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.
- 3DHGS — 3D-HGS: 3D Half-Gaussian Splatting (CVPR 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
- High-Quality Real-Time Rendering Using Subpixel Sampling Reconstruction (AAAI 2024 · similarity 0.84)
- GVKF: Gaussian Voxel Kernel Functions for Highly Efficient Surface Reconstruction in Open Scenes (NeurIPS 2024 · similarity 0.83)
- Micro-macro Wavelet-based Gaussian Splatting for 3D Reconstruction from Unconstrained Images (AAAI 2025 · similarity 0.83)
- Matérn Kernels for Tunable Implicit Surface Reconstruction (ICLR 2025 · similarity 0.83)
- Inverse Rendering using Multi-Bounce Path Tracing and Reservoir Sampling (ICLR 2025 · similarity 0.83)
- FreGS: 3D Gaussian Splatting with Progressive Frequency Regularization (CVPR 2024 · similarity 0.83)
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.
