ROI-GSurFisher: Next Best View Selection for Active Gaussian Splatting Via Fisher Information of ROI-Selected Gaussian Surfels
Wei Wang, Wei Ma, Hongliang Zhang, Hongbin Zha
AI summary
Problem
Existing Next Best View methods for Gaussian Splatting prioritize rendering uncertainty over geometric accuracy, often failing to select views that effectively reconstruct complex surfaces without requiring ground-truth data.
Approach
ROI-GSurFisher calculates Fisher Information gain specifically on 2D Gaussian surfels within a Region of Interest guided by surface geometry, then fuses this with a close-front view score that prioritizes nearby, surface-facing viewpoints.
Key results
- Superior geometric reconstruction quality compared to FisherRF at equal view counts
- SLAM systems achieve comparable mapping accuracy using only 200+ keyframes versus 800+ in baselines
- Novel uncertainty-based scoring module targeting geometric accuracy on complex surfaces
- Close-front view scoring module that prioritizes surface-facing viewpoints for better coverage
Why it matters
Enables robots and vision systems to build more accurate 3D maps faster by intelligently selecting geometrically informative views in Gaussian Splatting pipelines.
Abstract
Next Best View (NBV) selection is critical for achieving high-quality 3D reconstruction in unknown environ- ments. This paper presents an active NBV selection approach tailored for Gaussian Splatting (GS), a widely adopted 3D reconstruction technique that has recently gained significant attention and been extended to Simultaneous Localization and Mapping (SLAM) systems. Existing state-of-the-art NBV methods for GS focus on minimizing uncertainties of GS parameters but often fail to prioritize views that improve geometric reconstruction quality. To address this limitation, we propose an active view selection method for GS-based recon- struction, with its core being ROI-GSurFisher. This method calculates Fisher Information on Gaussian surfels selected via a Region of Interest (ROI) mechanism. Both the use of surfels for computation and the ROI constraint enhance ROI- GSurFisher’s ability to evaluate geometric information gain. We further introduce a close-front view scoring module that prioritizes viewpoints conducive to high-quality reconstruction. The final NBV is selected by maximizing the combined geomet- ric information gain and close-front score. Experimental results on 3D reconstruction of various objects and scenes demonstrate consistent qualitative and quantitative improvements. Beyond standalone 3D reconstruction, the proposed NBV method can be integrated into SLAM systems to select fewer but more valuable keyframes. Code is available at https://github. com/WW11111/ROI-GSurFisher.