Optimizing 3D Gaussian Splatting via Point Cloud Upsampling
For practitioners using 3DGS, this work provides preliminary guidelines on choosing upsampling strategies based on scene type, but the improvements are incremental and dataset-specific.
This paper evaluates point cloud upsampling methods (linear, triangular, spline, MLS, Voronoi) and a depth-guided lifting approach to improve 3D Gaussian Splatting initialization, achieving better reconstruction quality on Mip-NeRF360 and Replica datasets. The depth-guided method shows particular promise for texture-less regions.
3D Gaussian Splatting (3DGS) is a technique for creating and rendering 3D scenes, however its performance depends heavily on the quality of initial seed points. To improve 3DGS initialization, this study presents and evaluates several point cloud upsampling approaches: linear interpolation, triangular interpolation, spline-based surface reconstruction, moving least squares surface fitting, and Voronoi-based point generation. Additionally, this research introduces a depth-guided point lifting method that leverages depth maps to maintain geometric consistency with Structure-from-Motion (SfM) reconstructions. Through extensive experiments on the Mip-NeRF360 and Replica datasets, the proposed methods demonstrate improvements in reconstruction quality across diverse scene types. Results indicate that different upsampling strategies excel in different scenarios: surface reconstruction methods perform better with organic, detailed scenes, while simpler interpolation approaches are more suited for scenes dominated by piecewise-smooth geometries. In comparison, the depth-guided approach shows promise for adding geometry-aware points across the entire scene, importantly in texture-less regions. These findings, which provide preliminary practical guidelines for selecting appropriate upsampling methods based on scene characteristics and computational constraints, advances the understanding of how point cloud initialization affects 3DGS quality.