Optimized Minimal 4D Gaussian Splatting
Minseo Lee, Byeonghyeon Lee, Lucas Yunkyu Lee, Eunsoo Lee, Sangmin Kim, Seunghyeon Song, Joo Chan Lee, Jong Hwan Ko, Jaesik Park, Eunbyung Park
2025-10-07
Summary
This paper introduces a new technique called OMG4 for representing moving scenes in a way that's efficient for computers to render quickly. It's about making these scenes take up less storage space without losing visual quality.
What's the problem?
Representing dynamic, or moving, scenes with high detail requires a lot of data, specifically using a method called 4D Gaussian Splatting. This method uses millions of tiny 3D shapes (Gaussians) to build the scene, and storing all of them takes up a huge amount of memory. Previous attempts to reduce this storage requirement haven't been fully successful, either sacrificing how good the scene looks or not compressing the data enough.
What's the solution?
OMG4 tackles this problem by smartly reducing the number of Gaussians needed. It works in three steps: first, it identifies the most important Gaussians for accurately recreating the scene. Then, it removes Gaussians that are redundant or unnecessary. Finally, it combines similar Gaussians together. On top of that, it uses clever compression techniques to further shrink the size of the data needed to define how each Gaussian looks, making the whole representation much more compact.
Why it matters?
This research is important because it significantly reduces the storage space needed for complex moving scenes – by over 60% in their tests – while still maintaining a high level of visual quality. This makes it more practical to use these detailed scene representations in applications like virtual reality, augmented reality, and robotics where storage and processing power are limited, opening up possibilities for more realistic and interactive experiences.
Abstract
4D Gaussian Splatting has emerged as a new paradigm for dynamic scene representation, enabling real-time rendering of scenes with complex motions. However, it faces a major challenge of storage overhead, as millions of Gaussians are required for high-fidelity reconstruction. While several studies have attempted to alleviate this memory burden, they still face limitations in compression ratio or visual quality. In this work, we present OMG4 (Optimized Minimal 4D Gaussian Splatting), a framework that constructs a compact set of salient Gaussians capable of faithfully representing 4D Gaussian models. Our method progressively prunes Gaussians in three stages: (1) Gaussian Sampling to identify primitives critical to reconstruction fidelity, (2) Gaussian Pruning to remove redundancies, and (3) Gaussian Merging to fuse primitives with similar characteristics. In addition, we integrate implicit appearance compression and generalize Sub-Vector Quantization (SVQ) to 4D representations, further reducing storage while preserving quality. Extensive experiments on standard benchmark datasets demonstrate that OMG4 significantly outperforms recent state-of-the-art methods, reducing model sizes by over 60% while maintaining reconstruction quality. These results position OMG4 as a significant step forward in compact 4D scene representation, opening new possibilities for a wide range of applications. Our source code is available at https://minshirley.github.io/OMG4/.