发表机构
Case Western Reserve University（凯斯西储大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对3D高斯泼溅在分布外视角重建出现空洞和伪影的问题，本文提出独立视角检测与分层外推框架，并引入QA-Mask模块选择性利用生成数据，显著提升外推质量且可泛化。
AI 中文摘要
3D高斯泼溅在训练视角分布内实现了照片级逼真的重建，但在分布外的新视角上性能下降，在未观察区域出现空洞，在可观察区域出现伪影。近期工作将此任务表述为外推和内插，并尝试用生成模型解决，但在外推规模和质量上仍受限。它们重复“生成-重建-移位”循环以逐步构建场景，每一步都基于先前结果，从而引入累积误差。在本工作中，我们提出了一个用于外推和内插的整体框架。我们设计了一种独立的相机视角检测机制，以实现并行无冲突的外推，规避了对上述易出错循环的依赖。在此基础上，我们设计了一个分层流程，分别外推独立和依赖的相机视角。此外，先前方法忽略了生成图像与原始图像之间的不一致性，导致损害了重建良好的区域。我们提出了一个即插即用的质量感知掩码（QA-Mask）模块，能够选择性地利用生成数据。通过用像素级渲染质量校准学习权重，它防止了生成引起的对良好构建区域的退化。大量实验表明，我们的框架性能优越，且QA-Mask可泛化到多种生成式重建模型。
英文摘要
3D Gaussian Splatting achieves photorealistic reconstruction within training view distribution, yet it degrades on out-of-distribution novel views, exhibiting holes in unobserved regions and artifacts in observable areas. Recent works formulate this task as extrapolation and interpolation and try to address it with generative models, but remain limited in extrapolation scale and quality. They repeat a generate-reconstruct-shift cycle to progressively build a scene, which introduces accumulated errors with every step conditioning on previous outcomes. In this work, we propose a holistic framework for extrapolation and interpolation. We devise an independent camera view detection mechanism to enable parallel conflict-free extrapolation, circumventing the reliance on the aforementioned error-prone cycle. Building upon this, we design a hierarchical pipeline that extrapolates independent and dependent camera views separately. Additionally, previous methods overlook inconsistency between generated and original images, resulting in compromising well-reconstructed areas. We propose a plug-and-play Quality-Aware Mask (QA-Mask) module, enabling selective utilization of generated data. By calibrating learning weights with pixel-wise rendering quality, it prevents generation-induced degradations on well-constructed areas. Extensive experiments demonstrate the superior performance of our framework, with QA-Mask generalizing to multiple generative reconstruction models.
CommentsAccepted to ACM Multimedia (ACM MM) 2026. 9 pages
