Adaptive and Temporally Consistent Gaussian Surfels for Multi-view Dynamic Reconstruction

TL;DR

This paper introduces AT-GS, a novel method for high-quality, temporally consistent dynamic surface reconstruction from multi-view videos, effectively handling complex scenes with topology changes and rapid movements.

Contribution

The paper presents a per-frame incremental optimization approach with an adaptive densification strategy and curvature consistency to improve dynamic scene reconstruction.

Findings

Achieves superior accuracy in dynamic surface reconstruction

Reduces temporal jittering in reconstructed surfaces

Outperforms baseline methods on diverse datasets

Abstract

3D Gaussian Splatting has recently achieved notable success in novel view synthesis for dynamic scenes and geometry reconstruction in static scenes. Building on these advancements, early methods have been developed for dynamic surface reconstruction by globally optimizing entire sequences. However, reconstructing dynamic scenes with significant topology changes, emerging or disappearing objects, and rapid movements remains a substantial challenge, particularly for long sequences. To address these issues, we propose AT-GS, a novel method for reconstructing high-quality dynamic surfaces from multi-view videos through per-frame incremental optimization. To avoid local minima across frames, we introduce a unified and adaptive gradient-aware densification strategy that integrates the strengths of conventional cloning and splitting techniques. Additionally, we reduce temporal jittering in dynamic surfaces by ensuring consistency in curvature maps across consecutive frames. Our method achieves superior accuracy and temporal coherence in dynamic surface reconstruction, delivering high-fidelity space-time novel view synthesis, even in complex and challenging scenes. Extensive experiments on diverse multi-view video datasets demonstrate the effectiveness of our approach, showing clear advantages over baseline methods. Project page: \url{https://fraunhoferhhi.github.io/AT-GS}