PCAsplat: Gaussian Splatting with Local PCA Regularization
We introduce PCAsplat, a geometry-aware regularization framework for Gaussian splatting based on differentiable local principal component analysis (PCA).
ProofPaper ↗
Key points
- Gaussian splatting has emerged as a flexible representation for 3D reconstruction from posed images.
- We regularize the PCA eigenvalues to encourage Gaussians to move to the underlying surface with isotropic tangent-plane coverage.
- These surface-aligned splats enable downstream geometry-processing tasks, including point cloud segmentation, and direct Poisson reconstruction.
- Additionally, PCAsplat remains competitive under conventional novel view synthesis and mesh extraction tasks.
Sources (1)
- [1]PCAsplat: Gaussian Splatting with Local PCA RegularizationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 11:49 PM
We introduce PCAsplat, a geometry-aware regularization framework for Gaussian splatting based on differentiable local principal component analysis (PCA).
Gaussian splatting has emerged as a flexible representation for 3D reconstruction from posed images.
Extractive summary: sentences quoted from the sources.