SA-3DGS: A Self-Adaptive Compression Method for 3D Gaussian Splatting

TL;DR

SA-3DGS introduces a self-adaptive compression technique for 3D Gaussian Splatting that reduces storage costs significantly while preserving or enhancing rendering quality through importance scoring, clustering, and codebook repair.

Contribution

The paper presents a novel self-adaptive compression method that automatically identifies insignificant Gaussians, improves attribute compression, and repairs codebooks to maintain high-quality scene rendering.

Findings

Achieves up to 66x compression on benchmark datasets.

Maintains or improves rendering quality despite high compression ratios.

Enhances other pruning methods like LightGaussian with better performance.

Abstract

Recent advancements in 3D Gaussian Splatting have enhanced efficient and high-quality novel view synthesis. However, representing scenes requires a large number of Gaussian points, leading to high storage demands and limiting practical deployment. The latest methods facilitate the compression of Gaussian models but struggle to identify truly insignificant Gaussian points in the scene, leading to a decline in subsequent Gaussian pruning, compression quality, and rendering performance. To address this issue, we propose SA-3DGS, a method that significantly reduces storage costs while maintaining rendering quality. SA-3DGS learns an importance score to automatically identify the least significant Gaussians in scene reconstruction, thereby enabling effective pruning and redundancy reduction. Next, the importance-aware clustering module compresses Gaussians attributes more accurately into the codebook, improving the codebook's expressive capability while reducing model size. Finally, the codebook repair module leverages contextual scene information to repair the codebook, thereby recovering the original Gaussian point attributes and mitigating the degradation in rendering quality caused by information loss. Experimental results on several benchmark datasets show that our method achieves up to 66x compression while maintaining or even improving rendering quality. The proposed Gaussian pruning approach is not only adaptable to but also improves other pruning-based methods (e.g., LightGaussian), showcasing excellent performance and strong generalization ability.