LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS

Authors: Zhiwen Fan, Kevin Wang, Kairun Wen, Zehao Zhu, Dejia Xu, Zhangyang Wang

Submitted to arXiv on 28 Nov. 2023
Abstract

Recent advancements in real-time neural rendering using point-based techniques have paved the way for the widespread adoption of 3D representations. However, foundational approaches like 3D Gaussian Splatting come with a substantial storage overhead caused by growing the SfM points to millions, often demanding gigabyte-level disk space for a single unbounded scene, posing significant scalability challenges and hindering the splatting efficiency. To address this challenge, we introduce LightGaussian, a novel method designed to transform 3D Gaussians into a more efficient and compact format. Drawing inspiration from the concept of Network Pruning, LightGaussian identifies Gaussians that are insignificant in contributing to the scene reconstruction and adopts a pruning and recovery process, effectively reducing redundancy in Gaussian counts while preserving visual effects. Additionally, LightGaussian employs distillation and pseudo-view augmentation to distill spherical harmonics to a lower degree, allowing knowledge transfer to more compact representations while maintaining reflectance. Furthermore, we propose a hybrid scheme, VecTree Quantization, to quantize all attributes, resulting in lower bitwidth representations with minimal accuracy losses. In summary, LightGaussian achieves an averaged compression rate over 15x while boosting the FPS from 139 to 215, enabling an efficient representation of complex scenes on Mip-NeRF 360, Tank and Temple datasets. Project website: https://lightgaussian.github.io/

The papers around it, by date and closeness

Higher is closer to this paper; left is older. Lines join papers close to each other: groups of lines are sub-topics. Select a point to see why it is there. Back to the summary of this paper

How to read this map? Illustrated guide

Closest papers Prior work it builds on Papers that cite it

Papers around this one

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

  1. Very close Another approach to the problem
    EAGLES: Efficient Accelerated 3D Gaussians with Lightweight EncodingS
    Sharath Girish, Kamal Gupta, Abhinav Shrivastava · 2023 · Computer Vision and Pattern Recognition · cited 300 times · AI summary available

    EAGLES reduces 3D Gaussian storage with quantized embeddings and coarse-to-fine training, addressing the storage burden LightGaussian targets through different techniques.

    Abstract

    Recently, 3D Gaussian splatting (3D-GS) has gained popularity in novel-view scene synthesis. It addresses the challenges of lengthy training times and slow rendering speeds associated with Neural Radiance Fields (NeRFs). Through rapid, differentiable rasterization of 3D Gaussians, 3D-GS achieves real-time rendering and accelerated training. They, however, demand substantial memory resources for both training and storage, as they require millions of Gaussians in their point cloud representation for each scene. We present a technique utilizing quantized embeddings to significantly reduce memory storage requirements and a coarse-to-fine training strategy for a faster and more stable optimizatio…

  2. Very close Another approach to the problem Cited by this paper
    Compact 3D Gaussian Representation for Radiance Field
    Joo Chan Lee, Daniel Rho, Xiangyu Sun et al. · 2023 · Computer Vision and Pattern Recognition · cited 569 times · AI summary available

    Compact 3D Gaussian Representation compresses 3DGS to reduce storage while retaining radiance-field rendering, addressing the same compact-representation problem.

    Abstract

    Neural Radiance Fields (NeRFs) have demonstrated remarkable potential in capturing complex 3D scenes with high fidelity. However, one persistent challenge that hinders the widespread adoption of NeRFs is the computational bottleneck due to the volumetric rendering. On the other hand, 3D Gaussian splatting (3DGS) has recently emerged as an alternative representation that leverages a 3D Gaussisan-based representation and adopts the rasterization pipeline to render the images rather than volumetric rendering, achieving very fast rendering speed and promising image quality. However, a significant drawback arises as 3DGS entails a substantial number of 3D Gaussians to maintain the high fidelity o…

  3. Very close Another approach to the problem
    Compact 3D Scene Representation via Self-Organizing Gaussian Grids
    Wieland Morgenstern, Florian Barthel, Anna Hilsmann et al. · 2023 · Computer Vision and Pattern Recognition · cited 226 times · AI summary available

    Self-Organizing Gaussian Grids reduce 3DGS storage by arranging parameters in a locally homogeneous 2D grid, rather than LightGaussian’s pruning, distillation, and quantization.

    Abstract

    3D Gaussian Splatting has recently emerged as a highly promising technique for modeling of static 3D scenes. In contrast to Neural Radiance Fields, it utilizes efficient rasterization allowing for very fast rendering at high-quality. However, the storage size is significantly higher, which hinders practical deployment, e.g. on resource constrained devices. In this paper, we introduce a compact scene representation organizing the parameters of 3D Gaussian Splatting (3DGS) into a 2D grid with local homogeneity, ensuring a drastic reduction in storage requirements without compromising visual quality during rendering. Central to our idea is the explicit exploitation of perceptual redundancies pr…

  4. Very close Another approach to the problem
    FlexGaussian: Flexible and Cost-Effective Training-Free Compression for 3D Gaussian Splatting
    Bo-Yuan Tian, Qin-Zheng Gao, Siran Xianyu et al. · 2025 · ACM Multimedia · cited 13 times · Not summarized yet

    FlexGaussian compresses 3DGS without retraining, addressing model size with a training-free approach rather than LightGaussian’s pruning and recovery process.

    Abstract

    3D Gaussian Splatting has emerged as a prominent technique for representing and rendering complex 3D scenes, offering high fidelity and speed but resulting in large file sizes. Existing compression methods can reduce 3D Gaussian data size but often require costly retraining or refinement, which is memory- and compute-intensive and lacks flexibility for varying compression needs. This challenge grows as large-scale scenes become more common, increasing the demand for efficient, low-overhead compression methods - especially for resource-constrained mobile and edge devices. We introduce FlexGaussian, a flexible, cost-effective, training-free compression method for 3D Gaussian Splatting that com…

  5. Very close Another approach to the problem
    Optimized Minimal 3D Gaussian Splatting
    Joo Chan Lee, Jong Hwan Ko, Eunbyung Park · 2025 · Neural Information Processing Systems · cited 30 times · Not summarized yet

    Optimized Minimal 3D Gaussian Splatting reduces storage by using fewer Gaussians with high-precision attributes, addressing the same compression problem through a different trade-off.

    Abstract

    3D Gaussian Splatting (3DGS) has emerged as a powerful representation for real-time, high-performance rendering, enabling a wide range of applications. However, representing 3D scenes with numerous explicit Gaussian primitives imposes significant storage and memory overhead. Recent studies have shown that high-quality rendering can be achieved with a substantially reduced number of Gaussians when represented with high-precision attributes. Nevertheless, existing 3DGS compression methods still rely on a relatively large number of Gaussians, focusing primarily on attribute compression. This is because a smaller set of Gaussians becomes increasingly sensitive to lossy attribute compression, lea…

  6. Very close Another approach to the problem Cited by this paper
    Compressed 3D Gaussian Splatting for Accelerated Novel View Synthesis
    Simon Niedermayr, Josef Stumpfegger, Rüdiger Westermann · 2023 · Computer Vision and Pattern Recognition · cited 359 times · Not summarized yet

    Compressed 3D Gaussian Splatting uses sensitivity-aware vector clustering and quantization-aware training to compress Gaussian parameters and colors for novel-view synthesis.

    Abstract

    Recently, high-fidelity scene reconstruction with an optimized 3D Gaussian splat representation has been introducedfor novel view synthesis from sparse image sets. Making such representations suitable for applications like network streaming and rendering on low-power devices requires significantly reduced memory consumption as well as improved rendering efficiency. We propose a compressed 3D Gaussian splat representation that utilizes sensitivity-aware vector clustering with quantization-aware training to compress directional colors and Gaussian parameters. The learned codebooks have low bitrates and achieve a compression rate of up to 31 × on real-world scenes with only minimal degradation …

  7. Very close Groundwork for this paper Cited by this paper Used in depth by this paper A method it uses
    3D Gaussian Splatting for Real-Time Radiance Field Rendering
    Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler et al. · 2023 · ACM Transactions on Graphics, volume 42(4), July 2023 · cited 11,262 times · AI summary available

    3D Gaussian Splatting introduces the real-time radiance-field representation that LightGaussian compresses to reduce storage and improve rendering efficiency.

    Abstract

    Radiance Field methods have recently revolutionized novel-view synthesis of scenes captured with multiple photos or videos. However, achieving high visual quality still requires neural networks that are costly to train and render, while recent faster methods inevitably trade off speed for quality. For unbounded and complete scenes (rather than isolated objects) and 1080p resolution rendering, no current method can achieve real-time display rates. We introduce three key elements that allow us to achieve state-of-the-art visual quality while maintaining competitive training times and importantly allow high-quality real-time (>= 30 fps) novel-view synthesis at 1080p resolution. First, starting …

  8. Very close Another approach to the problem
    A Simple and Efficient 3D Gaussians Storage Format
    Denis Chaplygin, Juho Kannala, Esa Rahtu · 2026 · International Conference on Computer Vision Theory and Applications · Not summarized yet

    A Simple and Efficient 3D Gaussians Storage Format compresses Gaussian attributes using property-wise organization, delta coding, and quantization without scene-specific optimization.

    Abstract

    : Recent advances in 3D Gaussian Splatting (3DGS) have enabled photorealistic real–time rendering of complex scenes. However, the explicit storage of millions of Gaussian primitives leads to gigabyte–scale models, which hinders sharing, transmission, and deployment on resource-constrained devices. We propose a simple yet effective storage format tailored to 3DGS that exploits spatial redundancy through property-wise organization, delta coding, and lightweight quantization. Unlike generic compression schemes or learning-based methods, our approach requires no scene–specific optimization and introduces negligible computational overhead. Experiments on standard 3DGS benchmarks show that our for…

Prior work it builds on

  1. Very close Broader context A method it uses
    Scaffold-GS: Structured 3D Gaussians for View-Adaptive Rendering
    Tao Lu, Mulin Yu, Linning Xu et al. · 2023 · Computer Vision and Pattern Recognition · cited 1,052 times · AI summary available

    Scaffold-GS addresses redundant Gaussians through anchor-based structure for view-adaptive rendering, while LightGaussian targets storage reduction through pruning and compression.

    Abstract

    Neural rendering methods have significantly advanced photo-realistic 3D scene rendering in various academic and industrial applications. The recent 3D Gaussian Splatting method has achieved the state-of-the-art rendering quality and speed combining the benefits of both primitive-based representations and volumetric representations. However, it often leads to heavily redundant Gaussians that try to fit every training view, neglecting the underlying scene geometry. Consequently, the resulting model becomes less robust to significant view changes, texture-less area and lighting effects. We introduce Scaffold-GS, which uses anchor points to distribute local 3D Gaussians, and predicts their attri…

  2. Very close Broader context
    Flexible Techniques for Differentiable Rendering with 3D Gaussians
    L. Keselman, M. Hebert · 2023 · arXiv.org · preprint, not peer reviewed · cited 34 times · Not summarized yet

    Flexible Techniques extends differentiable Gaussian rendering for optical flow, mesh export, and normals, rather than compressing 3DGS storage.

    Abstract

    Fast, reliable shape reconstruction is an essential ingredient in many computer vision applications. Neural Radiance Fields demonstrated that photorealistic novel view synthesis is within reach, but was gated by performance requirements for fast reconstruction of real scenes and objects. Several recent approaches have built on alternative shape representations, in particular, 3D Gaussians. We develop extensions to these renderers, such as integrating differentiable optical flow, exporting watertight meshes and rendering per-ray normals. Additionally, we show how two of the recent methods are interoperable with each other. These reconstructions are quick, robust, and easily performed on GPU o…

  3. Close Broader context Used in depth by this paper A method it uses
    Compressing Volumetric Radiance Fields to 1 MB
    Lingzhi Li, Zhen Shen, Zhongshu Wang et al. · 2022 · Computer Vision and Pattern Recognition · cited 101 times · Not summarized yet

    Compressing Volumetric Radiance Fields to 1 MB compresses voxel-based radiance fields, addressing storage overhead in a different representation from LightGaussian’s 3D Gaussians.

    Abstract

    Approximating radiance fields with discretized volumetric grids is one of promising directions for improving NeRFs, represented by methods like DVGO, Plenoxels and TensoRF, which achieve super-fast training convergence and real-time rendering. However, these methods typically require a tremendous storage overhead, costing up to hundreds of megabytes of disk space and runtime memory for a single scene. We address this issue in this paper by introducing a simple yet effective framework, called vector quantized radiance fields (VQRF), for compressing these volume-grid-based radiance fields. We first present a robust and adaptive metric for estimating redundancy in grid models and performing vox…

  4. Close* Another approach to the problem
    Compact3D: Compressing Gaussian Splat Radiance Field Models with Vector Quantization
    K. Navaneet, Kossar Pourahmadi Meibodi, Soroush Abbasi Koohpayegani et al. · 2023 · arXiv.org · preprint, not peer reviewed · cited 111 times · Not summarized yet

    Compact3D applies vector quantization to compress Gaussian splat radiance-field models, addressing the same storage problem as LightGaussian.

  5. Close Broader context Used in depth by this paper A method it uses
    Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstruction
    Cheng Sun, Min Sun, Hwann-Tzong Chen · 2021 · Computer Vision and Pattern Recognition · cited 1,431 times · Not summarized yet

    Direct Voxel Grid Optimization reconstructs radiance fields with density and feature voxel grids, rather than compressing 3D Gaussian representations.

    Abstract

    We present a super-fast convergence approach to reconstructing the per-scene radiance field from a set of images that capture the scene with known poses. This task, which is often applied to novel view synthesis, is recently revolution-ized by Neural Radiance Field (NeRF) for its state-of-the-art quality and fiexibility. However, NeRF and its variants require a lengthy training time ranging from hours to days for a single scene. In contrast, our approach achieves NeRF-comparable quality and converges rapidly from scratch in less than 15 minutes with a single GPU. We adopt a representation consisting of a density voxel grid for scene geometry and a feature voxel grid with a shallow network fo…

How are these papers found?

We compare the title and abstract of this paper with those of every paper summarized on this site, and with the papers Semantic Scholar recommends, those it cites and those citing it (for a much-cited paper, its most recent citations and, from OpenAlex, the most cited ones). Among the prior work and the follow-ups, the papers that use each other in depth, a method it uses and the much-cited follow-ups come a little higher. The levels (very close, close, related) come from the similarity of the texts, not from a judgment of quality; * marks a paper without an abstract, compared by its title. The sentence under a paper is written by an AI from the two abstracts: check it in the papers. A paper marked not relevant by three readers leaves the list.

Updated on 4 November 2026.

The illustrated guide to the papers around a paper