EiGS: Event-Informed 3D Deblur Reconstruction with Gaussian Splatting
Yuchen Weng, Nuo Li, Peng Yu, qi Wang, Yongqiang Qi, Shaoze You, Jun Wang
AI summary
Problem
Standard 3D reconstruction methods like NeRF and 3D Gaussian Splatting fail under motion blur due to inaccurate pose estimation and inability to model complex camera jitter. Existing event-augmented approaches either suffer from slow rendering speeds or struggle with complex, non-linear camera motions.
Approach
EiGS integrates event streams with 3D Gaussian Splatting by using an Adaptive Deviation Estimator to predict per-Gaussian positional offsets that model camera shake. A motion consistency loss enforces global rigid motion coherence, while event and blurriness losses guide precise deblurring and optimization.
Key results
- Proposes EiGS, a framework combining event cameras and 3DGS for robust deblurred reconstruction
- Introduces a motion consistency loss that enforces globally coherent Gaussian displacements
- Develops an Adaptive Deviation Estimator network for precise per-Gaussian positional shift prediction
- Demonstrates superior sharpness and real-time rendering over NeRF and 3DGS baselines on real and synthetic datasets
Why it matters
Enables real-time, high-quality 3D scene recovery from motion-blurry inputs, advancing applications in robotics, autonomous driving, and AR/VR where camera instability is common.
Abstract
Neural Radiance Fields (NeRF) have significantly advanced photorealistic novel view synthesis. Recently, 3D Gaus- sian Splatting has emerged as a promising technique with faster training and rendering speeds. However, both methods rely heavily on clear images and precise camera poses, limiting performance under motion blur. To address this, we introduce Event-Informed 3D Deblur Reconstruction with Gaussian Splat- ting(EiGS), a novel approach leveraging event camera data to enhance 3D Gaussian Splatting, improving sharpness and clarity in scenes affected by motion blur. Our method employs an Adaptive Deviation Estimator to learn Gaussian center shifts as the inverse of complex camera jitter, enabling simulation of motion blur during training. A motion consistency loss ensures global coherence in Gaussian displacements, while Blurriness and Event Integration Losses guide the model toward precise 3D representations. Extensive experiments demonstrate superior sharpness and real-time rendering capabilities compared to existing methods, with ablation studies validating the effectiveness of our components in robust, high-quality reconstruction for complex static scenes.