发表机构
Zhejiang University（浙江大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
Bi-FlowGS通过光流连接生成式视图补全与高斯几何正则化，利用V2G和G2V双向流协同细化，缓解稀疏视角重建中的几何作弊问题，提升渲染质量与几何一致性。
AI 中文摘要
基于3D高斯泼溅（3DGS）的稀疏视角三维场景重建本质上是一个欠约束问题。合理的渲染结果也可能与错误的高斯几何共存，因为位置或深度上的误差可能被不透明度、尺度和外观所掩盖；我们将这种失效模式称为“几何作弊”。现有的正则化方法虽能约束几何，但仅限于已观测视角；而基于视频扩散的方法虽能补全未见视角，却主要将其用作RGB伪监督，未能充分利用运动和时间先验，且缺乏显式的几何监督。我们提出Bi-FlowGS，利用光流将生成式视图补全与高斯几何正则化相连接。我们即插即用的“视频到几何流蒸馏”（V2G）模块将修复后视频中的时间对应先验蒸馏到高斯几何中，以缓解几何作弊问题。相反，“几何到视频流引导修复”（G2V）模块利用当前3DGS几何引导时间一致的视频修复，提供更可靠的生成监督。V2G与G2V共同构成一个隐式的双向协同细化过程，使修复后的视频与优化的3DGS场景能够迭代地相互改进。实验表明，在宽基线和无边界360°基准上，渲染质量和几何一致性均得到提升。
英文摘要
Sparse-view 3D scene reconstruction with 3D Gaussian Splatting (3DGS) is inherently underconstrained. Plausible renderings can also coexist with erroneous Gaussian geometry, as errors in positions or depths may be concealed by opacity, scale, and appearance; we term this failure mode Geometry Cheating. Existing regularization methods constrain geometry but remain limited to observed views, while video-diffusion-based methods complete unseen views yet mainly use them as RGB pseudo-supervision, underusing motion and temporal priors and lacking explicit geometry supervision. We present Bi-FlowGS, which uses optical flow to bridge generative view completion and Gaussian geometry regularization. Our plug-and-play Video-to-Geometry Flow Distillation (V2G) distills temporal correspondence priors from restored videos into Gaussian geometry to alleviate Geometry Cheating. Conversely, Geometry-to-Video Flow-Guided Restoration (G2V) uses the current 3DGS geometry to guide temporally consistent video restoration, providing more reliable generative supervision. Together, V2G and G2V form an implicit bidirectional co-refinement process, enabling restored videos and the optimized 3DGS scene to iteratively improve each other. Experiments demonstrate improved rendering quality and geometric consistency across wide-baseline and unbounded 360° benchmarks.
