DeDiff-4DGS: Fusing Temporal Correlations and Diffusion Priors for Dynamic 3D Scenes
IEEE Transactions on Multimedia · 2026 · 28 · 7176-7188
DOI: 10.1109/tmm.2026.3673518 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Multimedia, 2026. Verfasst von Hoang Nguyen Nguyen und 6 weiteren Autor:innen. Die Arbeit nennt 59 Literaturangaben. Gefördert von: Australian Research Council Discovery.
Abstract (Originalsprache)
Reconstructing dynamic 3D (4D) scenes is challenging due to complex temporal dynamics and viewpoint sparsity in monocular videos. Existing extensions of 3D Gaussian Splatting (3D-GS) with its temporal modeling often fail to capture temporal correlations across frames, leading to redundant 3D Gaussians and reduced efficiency. To address this limitation, we propose DeDiff-4DGS, a framework that integrates temporal correlations and diffusion priors through two novel modules. The Temporal 3D Gaussian Latent Fusion (T3DLF) module fuses temporal information from sparse reference frames to promote spatio-temporal coherence and reduce the number of required 3D Gaussians. The Latent Diffusion Converter for 3D Gaussians (LDC3D) module enriches reference frames with semantic priors, complementing T3DLF under sparse-view conditions. Experimental results on standard benchmarks demonstrate that DeDiff-4DGS delivers higher reconstruction quality and improved efficiency over current state-of-the-art approaches.
Förderung
- Australian Research Council Discovery
Zitieren
APA
Nguyen, H. N., Xiang, W., Han, K., Yu, L., Lai, P., Chen, T., & Chen, Y. P. (2026). DeDiff-4DGS: Fusing Temporal Correlations and Diffusion Priors for Dynamic 3D Scenes. IEEE Transactions on Multimedia, 28, 7176-7188. https://doi.org/10.1109/tmm.2026.3673518
Vancouver
Nguyen HN, Xiang W, Han K, Yu L, Lai P, Chen T, et al. DeDiff-4DGS: Fusing Temporal Correlations and Diffusion Priors for Dynamic 3D Scenes. IEEE Transactions on Multimedia. 2026;28:7176-7188. doi:10.1109/tmm.2026.3673518
BibTeX
@article{Nguyen2026,
author = {Nguyen, Hoang Nguyen and Xiang, Wei and Han, Kang and Yu, Lu and Lai, Phu and Chen, Tianyu and Chen, Yi-Ping Phoebe},
title = {DeDiff-4DGS: Fusing Temporal Correlations and Diffusion Priors for Dynamic 3D Scenes},
journal = {IEEE Transactions on Multimedia},
year = {2026},
volume = {28},
pages = {7176--7188},
doi = {10.1109/tmm.2026.3673518}
}
EUVIMED