Research map / 3D vision / Gaussian splatting
MCGS: Multiview Consistency Enhancement for Sparse-View 3D Gaussian Radiance Fields
Yuru Xiao, Deming Zhai, Wenbo Zhao, Kui Jiang, Junjun Jiang, Xianming Liu
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 |
|---|---|---|---|
| add sparse initializer (A) to 3DGS (ablation) | 3DGS · LLFF (3 views) | PSNR: 19.09 -> 19.61 | Improved |
| add random filling (B) (ablation) | MCGS · LLFF (3 views) | PSNR: 19.61 -> 19.88 | Improved |
| add MVC-guided progressive pruning (C) (ablation) | MCGS · LLFF (3 views) | PSNR: 19.88 -> 20.29 | Improved |
| add edge-aware depth regularization (D) (ablation) | MCGS · LLFF (3 views) | PSNR: 20.29 -> 20.33 | Improved |
| use sparse initializer instead of COLMAP SFM (ablation) | 3DGS · LLFF (3 views) | PSNR: 19.09 -> 19.88 | Improved |
| use MCGS (pruning + filling) on random initialization (ablation) | 3DGS w/Random · LLFF (3 views, random init) | PSNR: 16.41 -> 19.29 | Improved |
| use COLMAP MVS dense initialization instead of sparse initializer (ablation) | MCGS · LLFF (3 views) | PSNR: 20.33 -> 20.42 | Mixed |
| account for volume assumption in feature sampling (ablation) | MCGS · LLFF (3 views) | PSNR: 20.33 -> 20.28 | Hurt |
| use constant threshold 0.90 instead of schedule (ablation) | MCGS · LLFF (3 views) | PSNR: 20.33 -> 19.55 | Hurt |
| use adaptive threshold instead of manual schedule (ablation) | MCGS · LLFF (3 views) | PSNR: 20.33 -> 20.26 | No clear effect |
| vary voxel resolution r from 16 to 128 (ablation) | MCGS · LLFF (3 views) | PSNR: 20.22 (r=16) to 20.36 (r=64) | No clear effect |
| replace LightGlue matcher with RoMa dense matcher (ablation) | MCGS · LLFF (3 views) | PSNR: 20.33 -> 20.62 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 4 on most shared results.
- DNGaussian — DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalization (CVPR 2024) · better on 3 of 3 shared results
- MVSplat — MVSplat: Efficient 3D Gaussian Splatting from Sparse Multi-View Images (ECCV 2024) · better on 2 of 3 shared results
- FreeNeRF — FreeNeRF: Improving Few-Shot Neural Rendering With Free Frequency Regularization (CVPR 2023) · better on 1 of 3 shared results
- MixNeRF — MixNeRF: Modeling a Ray With Mixture Density for Novel View Synthesis From Sparse Inputs (CVPR 2023) · 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
- Quantifying and Alleviating Co-Adaptation in Sparse-View 3D Gaussian Splatting (NeurIPS 2025 · similarity 0.92)
- MS-GS: Multi-Appearance Sparse-View 3D Gaussian Splatting in the Wild (NeurIPS 2025 · similarity 0.92)
- FewViewGS: Gaussian Splatting with Few View Matching and Multi-stage Training (NeurIPS 2024 · similarity 0.92)
- 3DGS-Enhancer: Enhancing Unbounded 3D Gaussian Splatting with View-consistent 2D Diffusion Priors (NeurIPS 2024 · similarity 0.92)
- Structure Consistent Gaussian Splatting with Matching Prior for Few-shot Novel View Synthesis (NeurIPS 2024 · similarity 0.91)
- Generative Sparse-View Gaussian Splatting (CVPR 2025 · similarity 0.91)
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.
