FeatureGS: Eigenvalue-Feature Optimization in 3D Gaussian Splatting for Geometrically Accurate and Artifact-Reduced Reconstruction
Miriam Jäger, Markus Hillemann, Boris Jutzi · 2025
Read FeatureGS: Eigenvalue-Feature Optimization in 3D Gaussian Splatting for Geometrically Accurate and Artifact-Reduced Reconstruction by Miriam Jäger, Markus Hillemann, Boris Jutzi (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-17655.txt.
Description
Miriam Jäger, Markus Hillemann, and Boris Jutzi add a geometry term to 3D Gaussian splatting so its blobs hug real surfaces, in this 2025 paper. Splatting rebuilds scenes from soft three-dimensional blobs, but blob centers drift off the true surface and stray floaters bloat the model. Their method scores local shape from eigenvalues and folds that score into training.
Four flavors of the score were tested on fifteen benchmark scenes. The best lifted geometric accuracy by thirty percent, cut blob counts by ninety percent, and cleared floaters while keeping picture quality. The authors target direct use of blob centers as geometry.
Catalog record
| CREATOR | Miriam Jäger, Markus Hillemann, Boris Jutzi |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-17655