Improving 3D Gaussian Splatting Compression by Scene-Adaptive Lattice Vector Quantization

2025-09-16

AI Summary

3D Gaussian Splatting (3DGS) enables high-fidelity, real-time rendering but suffers from prohibitively large raw data volumes, necessitating efficient compression. Existing approaches predominantly rely on uniform scalar quantization (USQ), which exhibits limited compression efficiency and lacks bitrate flexibility. This paper introduces lattice vector quantization (LVQ) to 3DGS compression for the first time, proposing Scene-Adaptive Lattice Vector Quantization (SALVQ). SALVQ jointly optimizes scene-specific lattice bases and scaling factors to achieve an optimal trade-off between rate–distortion (R–D) performance and computational complexity, while enabling dynamic, single-model bitrate adaptation and seamless integration into existing neural compression frameworks. Experiments demonstrate that SALVQ achieves average PSNR gains of 1.2–2.8 dB over USQ at equivalent bitrates, significantly outperforms USQ and leading learned compression methods in compression ratio, and reduces training time and GPU memory consumption by over 40%.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Learning on the Edge & Model Compression
• Data Mining & Knowledge Management: Data Compression

Application Category

• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

3D Gaussian Splatting (3DGS) is rapidly gaining popularity for its photorealistic rendering quality and real-time performance, but it generates massive amounts of data. Hence compressing 3DGS data is necessary for the cost effectiveness of 3DGS models. Recently, several anchor-based neural compression methods have been proposed, achieving good 3DGS compression performance. However, they all rely on uniform scalar quantization (USQ) due to its simplicity. A tantalizing question is whether more sophisticated quantizers can improve the current 3DGS compression methods with very little extra overhead and minimal change to the system. The answer is yes by replacing USQ with lattice vector quantization (LVQ). To better capture scene-specific characteristics, we optimize the lattice basis for each scene, improving LVQ's adaptability and R-D efficiency. This scene-adaptive LVQ (SALVQ) strikes a balance between the R-D efficiency of vector quantization and the low complexity of USQ. SALVQ can be seamlessly integrated into existing 3DGS compression architectures, enhancing their R-D performance with minimal modifications and computational overhead. Moreover, by scaling the lattice basis vectors, SALVQ can dynamically adjust lattice density, enabling a single model to accommodate multiple bit rate targets. This flexibility eliminates the need to train separate models for different compression levels, significantly reducing training time and memory consumption.

Problem

• Compressing 3DGS data to reduce storage and cost
• Replacing uniform scalar quantization with more efficient methods
• Enabling single-model compression for multiple bit rates

Innovation

• Scene-adaptive lattice vector quantization
• Dynamic lattice density scaling
• Seamless integration with existing architectures