Gong, JiarongUnger, JonasMiandji, EhsanGerrits, TimTeschner, Matthias2026-04-212026-04-212026978-3-03868-300-11017-4656https://doi.org/10.2312/egp.20261016https://diglib.eg.org/handle/10.2312/egp202610163D Gaussian Splatting (3DGS) suffers from large memory footprints. Existing compression techniques often lead to architectures with several additional trainable parameters and noticeable drops in rendering performance. We introduce the first dictionary-learning-based compression framework for 3DGS. Our compression framework is straightforward to implement, yet provides significant compression capabilities, preserves image quality, and improves real-time rendering performance. Across 13 benchmark scenes, our approach achieves an average compression ratio of 3.95×, 3.10×, and 4.55× when applied to 3DGS, 3DGS-MCMC, and PixelGS, respectively. This yields consistent rendering speedups of 23.3%, 24.3%, and 25.3%, while maintaining image quality.Attribution 4.0 International LicenseCCS Concepts: Computing methodologies → Rendering; Image-based renderingCCS ConceptsComputing methodologiesRenderingImage-based rendering10.2312/egp.202610162 pages