Rate distortion optimization improves compression of 3d gaussian splatting
RDCompGS: Rate-Distortion Optimization for Compression of 3D Gaussian Splatting with Image Codecs
Computer Vision and Pattern Recognition
Summary
Storing and sharing 3D images made of tiny glowing blobs called gaussians needs a lot of space. The authors found a way to better compress these 3D images by predicting how much space each part will take when using common image compressors. Their method carefully balances quality and size so the final 3D image looks good but uses less data. Tests show their approach saves a good amount of space compared to older methods.
What this means in practice
- •For 3d graphics developers: Compress 3D Gaussian Splatting models more efficiently for real-time rendering with less bandwidth.
- •For computer vision engineers: Reduce storage and transmission costs when sharing 3D reconstructions by optimizing compression rate and quality.
Authors
Huiping Ren, Shuai Wan, Luqian Bai, Zhang Chen, Fuzheng Yang
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.