4D Gaussian Splatting compression achieves over 40 times smaller files
TETRIS-4DGS: Efficient Post-Training Compression of 4D Gaussian Splatting
Computer Vision and Pattern Recognition
Summary
4D Gaussian Splatting is a way to create detailed, dynamic 3D scenes that can be seen in real time, but it creates huge files that are hard to store and share. The authors found a way to make these files much smaller without losing much detail by using smart techniques to decide which parts of the scene to compress more and by reorganizing data for existing image and video tools. Their method works quickly after the scene is created, making it easier to use these detailed 3D scenes in practice. This could help get high-quality dynamic 3D visuals into apps and devices where space and speed matter.
What this means in practice
- •For graphics software developers: Create highly detailed dynamic 3D scenes that can be compressed efficiently for fast loading and storage in interactive applications.
- •For game engine teams: Integrate compressed 4D Gaussian Splatting data to enable rich dynamic environments with reduced storage and bandwidth requirements.
Authors
Chengzhen Wu, Kelei Liu, Sicheng Li, Lu Yu, Yiyi Liao
Abstract
4D Gaussian Splatting (4DGS) has emerged as a powerful dynamic scene representation, supporting real-time high quality rendering but suffers from prohibitive storage overhead. Existing methods typically attempt to solve this by treating compression as a per-scene optimization problem, limiting generalizability and applicability where low-latency compression is required. In this work, we leverage existing image and video codecs and propose a generalized post-training compression framework for 4DGS to address these issues. Specifically, we apply importance-based adaptive quantization to Gaussians according to their importance levels, effectively refining bitrate allocation while remaining fully compatible with existing codecs. To further improve compression efficiency, we utilize the statistical properties of 4DGS and propose motion- and importance-aware packing to reorganize attributes into codec-friendly 2D maps. Our approach can achieve over 40x compression rate in seconds while maintaining high quality, facilitating the practical deployment of 4DGS.