A Threefold Compression Strategy for Lightweight 3D Gaussian Splatting

3D Gaussian Splatting (3DGS) excels in novel view synthesis with its real-time rendering capabilities, yet its reliance on massive Gaussian primitives and high-order spherical harmonics (SH) coefficients leads to excessive memory consumption, severely limiting deployment on resource-constrained platforms such as mobile devices and embedded systems. To address this challenge, we propose a threefold optimization framework. First, we design a global importance scoring mechanism to quantify the contribution of each Gaussian primitive to scene reconstruction and view synthesis, enabling precise pruning of redundant or low-impact primitives. Second, for regions with weak view-dependent effects, we introduce an adaptive spherical harmonics order adjustment strategy to dynamically reduce SH coefficients while preserving color fidelity, thereby minimizing computational and storage overhead. Third, we implement discrete quantization for low-sensitivity attributes using K-means clustering-based code-book construction combined with half-precision floating-point representation. Experiments demonstrate that our approach achieves an average compression rate exceeding <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$30 \times$</tex> across multiple standard scenes while more than doubling the rendering speed, effectively addressing memory bottlenecks and computational inefficiencies. This work paves the way for lightweight deployment of 3DGS in practical applications.