Research Paper#Image Compression, 2D Gaussian Splatting, Computer Vision🔬 ResearchAnalyzed: Jan 3, 2026 18:21
Structure-Guided 2D Gaussian Splatting for Image Compression
Published:Dec 30, 2025 06:35
•1 min read
•ArXiv
Analysis
This paper addresses the limitations of 2D Gaussian Splatting (2DGS) for image compression, particularly at low bitrates. It introduces a structure-guided allocation principle that improves rate-distortion (RD) efficiency by coupling image structure with representation capacity and quantization precision. The proposed methods include structure-guided initialization, adaptive bitwidth quantization, and geometry-consistent regularization, all aimed at enhancing the performance of 2DGS while maintaining fast decoding speeds.
Key Takeaways
- •Proposes a structure-guided allocation principle for 2D Gaussian Splatting to improve image compression.
- •Introduces structure-guided initialization, adaptive bitwidth quantization, and geometry-consistent regularization.
- •Achieves significant improvements in rate-distortion performance while maintaining fast decoding speeds (over 1000 FPS).
- •Demonstrates substantial BD-rate reduction compared to baseline methods.
Reference
“The approach substantially improves both the representational power and the RD performance of 2DGS while maintaining over 1000 FPS decoding. Compared with the baseline GSImage, we reduce BD-rate by 43.44% on Kodak and 29.91% on DIV2K.”