SplatGuide: Geometric Priors from 3D Gaussians for Pose-Free Novel View Synthesis
2026-08-17 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors present SplatGuide, a method to create new realistic views of a scene from unposed images without knowing the exact camera positions. Their approach improves on prior work by using a single 3D geometric reconstruction in three ways: rendering images for geometric guidance, using visibility maps to handle occlusions, and providing learned features for better synthesis. This combined use of information helps generate more accurate novel views. They tested SplatGuide on several datasets and showed it outperforms previous methods, even beating models that know the true camera poses in some cases.
novel view synthesis3D geometric reconstructionocclusion-aware reference selectionmulti-view diffusioncross-attentionrenderingpose-freeRealEstate10KNeRF
Authors
Yejun Zhang, Zihan Wang, Xu Ji, Yihao Wang, Yuxin Hou, Junyuan Fang, Juho-Matti Kilpeläinen, Arno Solin, Hamed Rezazadegan Tavakoli, Esa Rahtu, Juho Kannala
Abstract
Generating photorealistic novel views from unposed images requires both 3D geometric understanding and the ability to synthesize unseen content. A natural strategy combines feed-forward 3DGS reconstruction with multi-view diffusion. Yet prior pipelines extract at most one signal from the reconstruction, either pixel rendering or learned features, while none exploits per-Gaussian visibility for occlusion-aware reference selection. This *information disconnect* leaves renderable geometry, visibility cues, and learned features unused. SplatGuide closes this disconnect by reusing a single 3DGS scene across three complementary roles. Rendered images provide pixel-aligned geometric conditioning. Per-Gaussian source-view indices are rendered into a target-view voting map for occlusion-aware reference selection. Reconstruction tokens supply feature-level guidance via cross-attention. All three signals derive from the same reconstruction forward pass. Across RealEstate10K, DL3DV, Tanks-and-Temples, and Mip-NeRF 360, SplatGuide achieves state-of-the-art pose-free novel view synthesis. On RealEstate10K, with a moderate number of input views, it surpasses the ground-truth-pose baseline.