Temporally Aware Densification for Dynamic 3D Gaussian Splatting
2026-06-22 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors point out that current dynamic 3D Gaussian Splatting methods don't handle time changes well because they treat all parts statically, causing blurry results in moving areas. To fix this, they introduce a Visibility-Aware Densification system that considers when parts are actually visible over time. They also add a way to adjust how much each part is refined based on how long it exists, and a method to better handle fast movements by extending lifetimes of certain components. Overall, their approach improves clarity in dynamic scenes and works well as an add-on for existing methods.
3D Gaussian Splattingdynamic scenestemporal visibilitydensificationtemporally-adaptive thresholdingtemporal offset warpingmulti-view reconstructiondynamic reconstructionvisual quality
Authors
Vikram Sandu, Mayurdeep Pathak, Rajiv Soundararajan
Abstract
Despite modeling temporal motion, dynamic 3D Gaussian Splatting (3DGS) methods still inherit a static densification strategy that is ill-suited for dynamic scenes. This neglect of temporal behavior leads to under-reconstructed and blurry dynamic regions, as short-lived Gaussians receive sparse supervision and fail to densify effectively. We propose a Visibility-Aware Densification (VAD) framework that integrates temporal visibility into the densification process, ensuring that Gaussians are refined based on their actual temporal presence. A Temporally-Adaptive Thresholding (TAT) mechanism further adjusts each Gaussian's densification threshold according to its temporal lifespan, promoting balanced refinement of both static and dynamic regions. Finally, a Temporal Offset Warping (TOW) design enhances deformation capacity around temporal centers, extending the lifespan of highly dynamic Gaussians and facilitating more effective densification. Our approach achieves substantial improvements in the visual quality of dynamic regions, outperforming existing methods across three dynamic multi-view benchmark datasets. Moreover, the proposed VAD module generalizes across diverse dynamic 3DGS methods, consistently improving dynamic reconstruction as a plug-and-play component.