3D Gaussian Splatting (3DGS) has emerged as a dominant novel-view synthesis technique, but its high memory consumption severely limits its applicability on edge devices. A growing number of 3DGS compression methods have been proposed to make 3DGS more efficient, yet most only focus on storage compression and fail to address the critical bottleneck of rendering memory.
To address this problem, we introduce MEGS², a novel memory-efficient framework that tackles this challenge by jointly optimizing two key factors: the total primitive number and the parameters per primitive, achieving unprecedented memory compression. Specifically, we replace the memory-intensive spherical harmonics with lightweight arbitrarily-oriented spherical Gaussian lobes as our color representations. More importantly, we propose a unified soft pruning framework that models primitive-number and lobe-number pruning as a single constrained optimization problem.
Experiments show that MEGS² achieves a 50% static VRAM reduction and a 40% rendering VRAM reduction compared to existing methods, while maintaining comparable rendering quality.
The detailed architecture of our proposed MEGS² . (A) We first replace the memory-intensive Spherical Harmonics (SH) with Spherical Gaussian (SG) representation. (B) We formulate the compression as a memory-constrained optimization problem, which is solved using an ADMM-inspired approach that jointly adjusts primitive opacity and lobe sharpness to progressively sparsify both the number of primitives and the lobes. (C) Primitives and lobes with near-zero opacity and sharpness are removed, and a color compensation term is introduced to recover the energy of the removed lobes, thereby maintaining rendering quality with a significantly smaller memory consumption.
@misc{chen2025megs2memoryefficientgaussiansplatting,
title={MEGS$^{2}$: Memory-Efficient Gaussian Splatting via Spherical Gaussians and Unified Pruning},
author={Jiarui Chen and Yikeng Chen and Yingshuang Zou and Ye Huang and Peng Wang and Yuan Liu and Yujing Sun and Wenping Wang},
year={2025},
eprint={2509.07021},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2509.07021},
}