RDCompGS: Rate-Distortion Optimization for Compression of 3D Gaussian Splatting with Image Codecs

Huiping Ren; Shuai Wan; Luqian Bai; Zhang Chen; Fuzheng Yang

cs.CVOct 14, 2026

RDCompGS: Rate-Distortion Optimization for Compression of 3D Gaussian Splatting with Image Codecs

Authors: Huiping Ren, Shuai Wan, Luqian Bai, Zhang Chen, Fuzheng Yang

Organizations: School of Electronics and Information, Northwestern Polytechnical University, Xi’an 710129, China · Department of Computer Science, City University of Hong Kong, Hong Kong SAR, China · School of Telecommunication Engineering, Xidian University, Xi’an 710071, China

Abstract

3D Gaussian Splatting (3DGS) enables high-quality, real-time novel view synthesis but requires significant storage and transmission resources. Mapping Gaussian attributes to 2D representations enables efficient compression using mature image codecs. However, Gaussian representations that achieve comparable rendering quality are generally treated similarly during optimization, despite substantial differences in coding rate, leading to limited rate-distortion (R-D) performance after compression. To address this limitation, we propose RDCompGS, a 3DGS compression method that enables overall R-D optimization of Gaussian representations with respect to subsequent coding. A differentiable estimate of the coding rate is incorporated into the R-D optimization with rendering distortion. The rate constraint adjustment model further regulates the R-D trade-off across different compression conditions. Extensive experiments on three benchmark datasets demonstrate that RDCompGS achieves average Bjøntegaard Delta Rate reductions of 37.84% and 49.31% with H.265/HEVC and JPEG XL, respectively. The estimated rate also shows a strong correlation with the actual bit-rate under different compression conditions.

Figures & tables

Fig. 1: General pipeline for converting 3DGS into 2D representations and compressing them using image codecs.

Fig. 2: Overview of the proposed RDCompGS. To enable overall R-D optimization of Gaussian representations with respect to subsequent coding, RDCompGS first introduces differentiable rate estimation based on the distribution of transformed coefficients and their quantization probabilities. Then, the rate constraint adjustment model regulates the rate constraint across different compression conditions. Finally, the estimated rate and rendering distortion jointly guide Gaussian optimization for subsequent coding.

Fig. 3: Spatial correlation comparison between a natural image and Gaussian attribute grids on the Truck scene. (a) Natural image; (b) and (c) scale attribute grids without and with RDCompGS, respectively; (d), (e), and (f) spatial correlation maps of (a), (b), and (c), respectively. Spatial correlation is calculated within a $21\times 21$ window by averaging the horizontal and vertical Pearson correlations, with gray to blue indicating increasing correlation. The scale attribute shows lower spatial correlation than the natural image, while higher correlation is observed with RDCompGS.

Panel of Figure 4 — Fig. 4: Laplace distribution fitting of DCT coefficients for opacity and SH DC component during training on the Truck scene.

Fig. 5: R-D performance comparison on the Truck scene under H.265/HEVC compression between Gaussian optimization without the rate constraint and with the rate constraint using

\lambda_{\mathrm{rate}}=1\times 10^{-7}

at QP values of 4, 10, 16, and 22.

Dataset	Method	Distance = 0.1005				Distance = 0.2157				Distance = 0.4511				Distance = 0.9363				BD-Rate (%) $\downarrow$
Dataset	Method	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	BD-Rate (%) $\downarrow$
Tanks & Temples (Train)	SOG	14.828	21.54	0.794	0.231	11.343	21.36	0.791	0.238	7.779	20.78	0.778	0.250	6.896	20.25	0.766	0.257	-30.44
Tanks & Temples (Train)	RDCompGS	16.194	21.73	0.796	0.235	10.488	21.50	0.791	0.241	6.321	21.06	0.779	0.252	4.079	20.70	0.758	0.274	-30.44
Deep Blending	SOG	12.175	29.22	0.891	0.276	8.969	28.93	0.889	0.280	6.338	28.15	0.882	0.288	5.845	27.68	0.879	0.292	-60.94
Deep Blending	RDCompGS	11.623	29.52	0.898	0.271	7.185	29.32	0.895	0.273	4.562	29.20	0.889	0.282	3.136	28.63	0.879	0.299	-60.94
Mip-NeRF 360	SOG	37.140	26.82	0.797	0.235	28.136	26.53	0.793	0.244	19.073	25.62	0.776	0.261	16.933	24.88	0.762	0.271	-56.55

TABLE I: R-D performance of RDCompGS using JPEG XL compared with SOG. Size is measured in MB and PSNR in dB.

Dataset	Method	QP = 4				QP = 10				QP = 16				QP = 22				BD-Rate (%) $\downarrow$
Dataset	Method	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	BD-Rate (%) $\downarrow$
Tanks & Temples (Train)	SOG	14.828	21.54	0.794	0.231	11.343	21.36	0.791	0.238	7.779	20.78	0.778	0.250	6.896	20.25	0.766	0.257	-15.44
Tanks & Temples (Train)	RDCompGS	16.451	21.57	0.795	0.237	11.180	21.24	0.785	0.244	6.886	21.04	0.776	0.255	4.645	20.51	0.753	0.279	-15.44
Deep Blending	SOG	12.175	29.22	0.891	0.276	8.969	28.93	0.889	0.280	6.338	28.15	0.882	0.288	5.845	27.68	0.879	0.292	-50.48
Deep Blending	RDCompGS	11.554	29.46	0.897	0.271	7.603	29.25	0.895	0.275	4.846	29.04	0.888	0.283	3.417	28.36	0.876	0.307	-50.48
Mip-NeRF 360	SOG	37.140	26.82	0.797	0.235	28.136	26.53	0.793	0.244	19.073	25.62	0.776	0.261	16.933	24.88	0.762	0.271	-47.61

TABLE II: R-D performance of RDCompGS using H.265/HEVC compared with SOG. Size is measured in MB and PSNR in dB.

Fig. 6: R-D performance comparison between SOG and RDCompGS compressed using JPEG XL on three benchmark datasets in terms of PSNR, SSIM, and LPIPS.

Fig. 7: Visual comparison on the Train, Playroom, and Bonsai scenes, where RDCompGS is compressed using JPEG XL. Each row shows GT, 3DGS, SOG, and RDCompGS from left to right, with the same local region enlarged for comparison.

Fig. 8: Estimated rate and actual bit-rate produced by H.265/HEVC at different QP values. Each point corresponds to one Gaussian attribute image.

Fig. 9: R-D curves on the Truck scene for different

\lambda_{\mathrm{ref}}

values. The selected value

\lambda_{\mathrm{ref}}=1\times 10^{-7}

achieves a favorable R-D trade-off according to the observed trend of the curves.

Fig. 10: BD-Rate comparison of different

\beta

values on the Truck scene under H.265/HEVC compression.

Dataset	Method	QP = 4				QP = 10				QP = 16				QP = 22				BD-Rate (%) $\downarrow$
Dataset	Method	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	Size	PSNR	SSIM	LPIPS	BD-Rate (%) $\downarrow$
Tanks & Temples (Train)	LightGaussian	15.367	21.52	0.780	0.238	12.431	21.46	0.776	0.241	9.692	21.24	0.767	0.251	7.088	20.67	0.736	0.283	-38.83
Tanks & Temples (Train)	Ours	14.657	21.52	0.781	0.237	10.983	21.47	0.778	0.240	7.012	21.34	0.768	0.251	3.868	20.76	0.743	0.278	-38.83
Deep Blending	LightGaussian	36.044	29.76	0.898	0.256	29.350	29.72	0.897	0.257	23.069	29.57	0.894	0.262	17.042	29.06	0.883	0.277	-67.19
Deep Blending	Ours	25.096	29.95	0.902	0.254	13.073	29.86	0.901	0.257	7.861	29.63	0.897	0.265	5.731	28.96	0.885	0.285	-67.19
Mip-NeRF 360	LightGaussian	39.069	26.65	0.789	0.232	31.674	26.54	0.786	0.236	24.776	26.21	0.776	0.247	18.223	25.38	0.749	0.276	-54.54

TABLE III: R-D performance comparison with LightGaussian under H.265/HEVC compression. Here, Ours denotes RDCompGS with the pruning strategy proposed in LightGaussian. Size is measured in MB and PSNR in dB.

Dataset	BD-Rate (%) $\downarrow$	BD-PSNR (dB) $\uparrow$
Tanks & Temples (Train)	-53.28	0.63
Deep Blending	-62.32	0.88
Mip-NeRF 360	-56.09	1.18

TABLE IV: Ablation of differentiable rate estimation under H.265/HEVC compression, using Gaussian optimization without the estimated rate term as the reference.

Dataset	BD-Rate (%) $\downarrow$	BD-PSNR (dB) $\uparrow$
Tanks & Temples (Train)	0.77	0.00
Deep Blending	-0.59	0.02
Mip-NeRF 360	-5.46	0.25

TABLE V: Comparison of the rate constraint adjustment model with the use of a single rate constraint under H.265/HEVC compression, using the latter as the reference.

3DGS Compression

3D Gaussian Splatting compression reduces storage and rendering costs while preserving the quality of reconstructed views and scenes.

See all 80 papers

MesonGS++: Post-training Compression of 3D Gaussian Splatting with Hyperparameter Searching
Apr 29, 2026Shuzhao Xie, Junchen Ge, Weixiang Zhang +103DGS Compression 3D Gaussian Splatting
Shenzhen International Graduate School, Tsinghua University, Shenzhen, China · The Hong Kong University of Science and Technology, Hong Kong SAR, China · Harbin Institute of Technology, Shenzhen, China +6
Compressing 3D Gaussian Splatting via Cross-Representation Priors
Sep 19, 2026Yezheng Zhang, Huanxiong Liang, Chuqin Zhou +23DGS Compression Gaussian Splatting Compression
Institute of Image Communication and Network Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
Towards Practical Compression of 3D Gaussian Splatting
Sep 24, 2026Pengpeng Yu, Yueru Chen, Fei Song +43DGS Compression Entropy Coding
Sun Yat-sen University, China · Pengcheng Laboratory, China · Academy of Broadcasting Science, National Radio and Television Administration, China +1

Apr 29, 2026J/K move · Enter open · S save

MesonGS++: Post-training Compression of 3D Gaussian Splatting with Hyperparameter Searching

Shuzhao Xie, Junchen Ge, Weixiang Zhang, Jiahang Liu+9

Shenzhen International Graduate School, Tsinghua University, Shenzhen, China · The Hong Kong University of Science and Technology, Hong Kong SAR, China · Harbin Institute of Technology, Shenzhen, China+6

3DGS Compression 3D Gaussian Splatting 3D Gaussian Pruning Post-Training Quantization Gaussian Splatting Compression

Open paper AskPDF arXiv

3D Gaussian Splatting (3DGS) achieves high-quality novel view synthesis with real-time rendering, but its storage cost remains prohibitive for practical deployment. Existing post-training compression methods still rely on many coupled hyperparameters across pruning, transformation, quantization, and entropy coding, making it difficult to control the final compressed size and fully exploit the rate-distortion trade-off. We propose MesonGS++, a size-aware post-training codec for 3D Gaussian compression. On the codec side, MesonGS++ combines joint importance-based pruning, octree geometry coding, attribute transformation, selective vector quantization for higher-degree spherical harmonics, and group-wise mixed-precision quantization with entropy coding. On the configuration side, it treats the reserve ratio and bit-width allocation as the dominant rate-distortion knobs and jointly optimizes them under a target storage budget via discrete sampling and 0--1 integer linear programming. We further propose a linear size estimator and a CUDA parallel quantization operator to accelerate the hyperparameter searching process. Extensive experiments show that MesonGS++ achieves over 34

\times

compression while preserving rendering fidelity, outperforming state-of-the-art post-training methods and accurately meeting target size budgets. Remarkably, without any training, MesonGS++ can even surpass the PSNR of vanilla 3DGS at a 20

\times

compression rate on the Stump scene. Our code is available at https://github.com/mmlab-sigs/mesongs_plus

Abstract

3DGS Compression

Explore similar work

MesonGS++: Post-training Compression of 3D Gaussian Splatting with Hyperparameter Searching

Compressing 3D Gaussian Splatting via Cross-Representation Priors

Towards Practical Compression of 3D Gaussian Splatting