Breaking the Vicious Cycle: Coherent 3D Gaussian Splatting from Sparse and Motion-Blurred Views

2025-12-11

AI Summary

This work addresses the challenging problem of high-fidelity 3D reconstruction from sparse and motion-blurred images, where view sparsity and motion blur mutually exacerbate each other—leading to fragmented reconstructions and severe low-frequency bias in 3D Gaussian Splatting (3DGS). To tackle this, we propose a dual-prior collaborative framework that jointly leverages a pre-trained deblurring network (providing photometric priors) and a diffusion model (encoding geometric priors), augmented by a consistency-driven camera exploration module and a depth regularization loss. Our method achieves robust multi-view correspondence and improved geometric fidelity using only 3–9 sparse, blurred input views. Extensive experiments demonstrate state-of-the-art performance: our approach surpasses existing methods in both quantitative metrics (e.g., PSNR, SSIM, LPIPS) and qualitative visual quality for novel-view synthesis, establishing a new SOTA for this task.

Technology Category

• Computer Vision: 3D Computer Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

3D Gaussian Splatting (3DGS) has emerged as a state-of-the-art method for novel view synthesis. However, its performance heavily relies on dense, high-quality input imagery, an assumption that is often violated in real-world applications, where data is typically sparse and motion-blurred. These two issues create a vicious cycle: sparse views ignore the multi-view constraints necessary to resolve motion blur, while motion blur erases high-frequency details crucial for aligning the limited views. Thus, reconstruction often fails catastrophically, with fragmented views and a low-frequency bias. To break this cycle, we introduce CoherentGS, a novel framework for high-fidelity 3D reconstruction from sparse and blurry images. Our key insight is to address these compound degradations using a dual-prior strategy. Specifically, we combine two pre-trained generative models: a specialized deblurring network for restoring sharp details and providing photometric guidance, and a diffusion model that offers geometric priors to fill in unobserved regions of the scene. This dual-prior strategy is supported by several key techniques, including a consistency-guided camera exploration module that adaptively guides the generative process, and a depth regularization loss that ensures geometric plausibility. We evaluate CoherentGS through both quantitative and qualitative experiments on synthetic and real-world scenes, using as few as 3, 6, and 9 input views. Our results demonstrate that CoherentGS significantly outperforms existing methods, setting a new state-of-the-art for this challenging task. The code and video demos are available at https://potatobigroom.github.io/CoherentGS/.

Problem

• Reconstructs 3D scenes from sparse, blurry images
• Breaks the cycle where blur and sparsity degrade reconstruction
• Uses dual generative priors for detail and geometry

Innovation

• Uses dual-prior strategy with deblurring and diffusion models
• Integrates consistency-guided camera exploration for adaptive generation
• Applies depth regularization loss to ensure geometric plausibility