3DGS.zip: A survey on 3D Gaussian Splatting Compression Methods

TL;DR

The paper surveys compression and compaction techniques for 3D Gaussian Splatting (3DGS), an explicit radiance-field representation that trades storage for real-time rendering. It categorizes approaches into attribute-driven compression and structure-based compaction, detailing methods from vector quantization, anchors, hash grids, Z-ordering, and tri-planes to adaptive density control and pruning. Through a unified evaluation framework, it compares methods across standard datasets (Tanks and Temples, Mip-NeRF 360, Deep Blending, Synthetic NeRF) using PSNR, SSIM, LPIPS, and model size, highlighting that context-aware and hierarchical schemes often balance quality and size best, while aggressive compression methods excel in extreme storage reduction. The survey also outlines practical guidelines, dataset/testing protocols, and promising future directions, such as dynamic scenes, level-of-detail representations, and cross-scene codebooks, to broaden 3DGS applicability. A community resource site is maintained for ongoing updates and standardized benchmarking.

Abstract

3D Gaussian Splatting (3DGS) has emerged as a cutting-edge technique for real-time radiance field rendering, offering state-of-the-art performance in terms of both quality and speed. 3DGS models a scene as a collection of three-dimensional Gaussians, with additional attributes optimized to conform to the scene's geometric and visual properties. Despite its advantages in rendering speed and image fidelity, 3DGS is limited by its significant storage and memory demands. These high demands make 3DGS impractical for mobile devices or headsets, reducing its applicability in important areas of computer graphics. To address these challenges and advance the practicality of 3DGS, this survey provides a comprehensive and detailed examination of compression and compaction techniques developed to make 3DGS more efficient. We classify existing methods into two categories: compression, which focuses on reducing file size, and compaction, which aims to minimize the number of Gaussians. Both methods aim to maintain or improve quality, each by minimizing its respective attribute: file size for compression and Gaussian count for compaction. We introduce the basic mathematical concepts underlying the analyzed methods, as well as key implementation details and design choices. Our report thoroughly discusses similarities and differences among the methods, as well as their respective advantages and disadvantages. We establish a consistent framework for comparing the surveyed methods based on key performance metrics and datasets. Specifically, since these methods have been developed in parallel and over a short period of time, currently, no comprehensive comparison exists. This survey, for the first time, presents a unified framework to evaluate 3DGS compression techniques. We maintain a website that will be regularly updated with emerging methods: https://w-m.github.io/3dgs-compression-survey/ .

Figures (20)

• Figure 1: Histograms of the opacity and the first scaling attributes for all 3D Gaussians of the bicycle scene (as trained by 3DGSkerbl3Dgaussians).
• Figure 2: A correlation heatmap for attributes of all 3D Gaussians of the bicycle scene (as trained by 3DGSkerbl3Dgaussians).
• Figure 3: Vector Quantization steps.
• Figure 4: Example of a point cloud with Octree partitioning.
• Figure 5: Anchor-based structure. Left: A voxalized scene with SfM inistialized points. Right: The center of each voxel becomes an anchor and is associated with position, feature, scaling and offset. From each anchor neural Gaussians are spawned.