MesonGS: Post-training Compression of 3D Gaussians via Efficient Attribute Transformation ECCV 2024
- Shuzhao Xie
- Weixiang Zhang
- Chen Tang
- Yunpeng Bai
- Rongwei Lu
- Shijia Ge
- Zhi Wang Tsinghua University, Peng Cheng Laboratory The Chinese University of Hong Kong, The University of Texas at Austin
Abstract
3D Gaussian Splatting demonstrates excellent quality and speed in novel view synthesis. Nevertheless, the significant size of the 3D Gaussians presents challenges for transmission and storage. Current approaches employ compact models to compress the substantial volume and attributes of 3D Gaussians, along with intensive training to uphold quality. These endeavors demand considerable finetuning time, presenting formidable hurdles for practical deployment. To this end, we propose MesonGS, a codec for post-training compression of 3D Gaussians. Initially, we introduce a measurement criterion that considers both view-dependent and view-independent factors to assess the impact of each Gaussian point on the rendering output, enabling the removal of insignificant points. Subsequently, we decrease the entropy of attributes through two transformations that complement subsequent entropy coding techniques to enhance the file compression rate. More specifically, we first replace the rotation quaternion with Euler angles; then, we apply region adaptive hierarchical transform (RAHT) to key attributes to reduce entropy. Lastly, we suggest block quantization to control quantization granularity, thereby avoiding excessive information loss caused by quantization. Moreover, a finetune scheme is introduced to restore quality. Extensive experiments demonstrate that MesonGS significantly reduces the size of 3D Gaussians while preserving competitive quality.
Method
|
1) We prune insignificant Gaussians by considering both view-dependent and view-independent factors. 2) Geometry compression is performed using an octree to generate voxelized coordinates for future transformations. 3) We replace rotation quaternions with Euler angles. 4) Applying RAHT and 5) block quantization to important attributes. Notably, RAHT is not applied to the scales when the quantization bit is 8 (see dashed arrow). 6) To significantly compress the remained SH coefficients, vector quantization is employed. 7) All components are packed by zip.
Quantitative Results
Check the wonderful 3DGS compression survey.
Video demo
Novel view reconstructions for 3DGS (left 386MB) and our method (right 15MB). Our method can achieve 25x compression rate while maintain similar visual quality.
Citation
Acknowledgements
This work is supported in part by National Key Research and Development Project of China (Grant No. 2023YFF0905502), Shenzhen Science and Technology Program (Grant No. JCYJ20220818101014030). We thank the valuable advices from anonymous reviewers. We thank JiangXingAI for sponsoring the research.
The website template was borrowed from Michaël Gharbi, Ref-NeRF, and Zip-NeRF.