HybridGS: High-Efficiency Gaussian Splatting Data Compression using Dual-Channel Sparse Representation and Point Cloud Encoder

TL;DR

HybridGS addresses the data-volume challenge of 3D Gaussian Splatting (3DGS) by coupling a dual-channel, learnable sparse generation stage with a standard point cloud encoder (GPCC) to form output bitstreams that are fast to encode/decode and readily deployable. It introduces Learnable Quantizer-based Positioning (LQM) and latent-feature-based attribute compression to produce explicit, compact 3DGS, then leverages downstream GAN-free encoding with rate control via primitive pruning and BD adaptation. The approach yields reconstruction quality comparable to state-of-the-art generative compression methods while delivering substantially faster coding times and full compatibility with existing point-cloud codecs and pipelines. This work highlights practical deployment potential for 3DGS in streaming and real-time contexts, and points to future improvements in adaptive parameter selection and 3D-space loss considerations for further gains.

Abstract

Most existing 3D Gaussian Splatting (3DGS) compression schemes focus on producing compact 3DGS representation via implicit data embedding. They have long coding times and highly customized data format, making it difficult for widespread deployment. This paper presents a new 3DGS compression framework called HybridGS, which takes advantage of both compact generation and standardized point cloud data encoding. HybridGS first generates compact and explicit 3DGS data. A dual-channel sparse representation is introduced to supervise the primitive position and feature bit depth. It then utilizes a canonical point cloud encoder to perform further data compression and form standard output bitstreams. A simple and effective rate control scheme is proposed to pivot the interpretable data compression scheme. At the current stage, HybridGS does not include any modules aimed at improving 3DGS quality during generation. But experiment results show that it still provides comparable reconstruction performance against state-of-the-art methods, with evidently higher encoding and decoding speed. The code is publicly available at https://github.com/Qi-Yangsjtu/HybridGS.

Figures (15)

• Figure 1: Existing generative compression frameworks and the proposed HybridGS.
• Figure 2: Framework of HybridGS.
• Figure 3: Robust Quantization of 3DGS features.
• Figure 4: Scheme of LQM.
• Figure 5: RD curve.

Theorems & Definitions (1)

• proof