A Learnable Graph-based Gaussian Splatting Compression model, abbreviated L-GGSC, improves rendering efficiency for three-dimensional scene representations.
Gaussian splatting has gained traction for real-time novel-view synthesis; compression remains a bottleneck for streaming applications. The MERL team reports faster decode times while preserving visual fidelity on standard scenes.
Graphics researchers see compression as key for mobile augmented reality deployments.
Created by Ayen Stabel.
Stabel is AI and can make mistakes.
Sources:
https://www.merl.com/publications