3DGS-VBench+ and 3DGSV-Assessor: A Multi-Dimensional Subjective Benchmark and a Reference-Flexible Objective Model for 3D Gaussian Splatting Video Quality Assessment

📄 arXiv: 2611.04498v1 📥 PDF

作者: Yuke Xing, William Gordon, Jiarui Wang, Yujie Zhang, Guangtao Zhai, Yiling Xu

分类: cs.CV

发布日期: 2026-11-03

🔗 代码/项目: GITHUB


💡 一句话要点

提出3DGS-VBench+和3DGSV-Assessor以解决3D高斯点视频质量评估问题

🎯 匹配领域: 支柱三:空间感知与语义 (Perception & Semantics) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 视频质量评估 3D高斯点 多维度评估 生成模型 流媒体技术

📋 核心要点

  1. 现有的3D高斯点视频质量评估方法在处理几何和颜色失真时存在维度特定的缺陷,缺乏有效的多维度评估机制。
  2. 本文提出3DGS-VBench+作为主观视频质量评估基准,并开发3DGSV-Assessor框架,支持多维度质量评分和灵活的参考机制。
  3. 实验表明,3DGSV-Assessor在与人类感知的相关性上表现优异,并在其他新视角合成基准上展现出强大的泛化能力。

📝 摘要(中文)

3D高斯点(3DGS)技术支持实时新视角合成(NVS),但其存储需求高,压缩变得至关重要。现有生成压缩方法在处理几何和颜色属性时采用不同参数,导致维度特定的失真,亟需多维度评估。由于现实世界捕获的合成视频缺乏对齐的真实视频作为参考,本文提出3DGS-VBench+,一个包含3036个3DGS重建视频的主观视频质量评估基准,并基于此开发了3DGSV-Assessor,一个灵活的多模态框架,支持减少参考和无参考评估。实验结果显示,该方法在与人类感知的相关性上达到最先进水平,并在其他NVS基准上具有强大的零-shot泛化能力。

🔬 方法详解

问题定义:本文旨在解决3D高斯点视频质量评估中的维度特定失真问题,现有方法在处理几何和颜色属性时未能有效整合,导致评估结果不准确。

核心思路:通过构建3DGS-VBench+基准和3DGSV-Assessor框架,本文提出了一种多维度质量评估方法,能够在缺乏对齐参考的情况下进行有效评估。

技术框架:整体架构包括数据集的构建和多模态评估框架,3DGS-VBench+包含3036个重建视频和9108个主观评分,3DGSV-Assessor则通过一个有界门将排名参考线索融合到回归器中。

关键创新:3DGSV-Assessor的最大创新在于其参考灵活性,能够同时支持减少参考和无参考的质量评估,与现有方法相比,显著提升了评估的准确性和适用性。

关键设计:在设计中,采用了多模态融合技术,结合了主观评分数据和生成模型的输出,损失函数设计上考虑了多维度的评分标准,以确保评估的全面性。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示,3DGSV-Assessor在3DGS-VBench+上与人类感知的相关性达到最先进水平,且在其他新视角合成基准上展现出强大的零-shot泛化能力,提升幅度显著,具体性能数据未明确给出。

🎯 应用场景

该研究的潜在应用领域包括视频编码、流媒体传输和虚拟现实等场景,能够为视频质量评估提供更准确的工具,提升用户体验。未来,该方法可能推动视频压缩技术的发展,促进更高效的内容分发和存储解决方案。

📄 摘要(原文)

3D Gaussian Splatting (3DGS) enables real-time novel view synthesis (NVS), but its heavy storage makes compression essential, and both training and compression introduce representation-specific distortions. Since generative compression in many methods treats geometric and color attributes via different parameters, it can induce dimension-specific distortions, motivating multi-dimensional assessment. Moreover, real-world captures generally lack an aligned ground-truth video for a synthesized, continuous camera trajectory, leaving only sparse captured views as references. To enable multi-dimensional quality assessment of 3DGS videos with scarce aligned references, we first present 3DGS-VBench+, a large-scale subjective video quality assessment (VQA) benchmark comprising 3{,}036 3DGS reconstructions, each paired with a rendered video, across 11 real-world and 22 synthetic scenes, generated by 6 representative compression algorithms at systematically designed levels and annotated with 9{,}108 MOSs covering overall, geometry, and color quality. Based on 3DGS-VBench+, we propose 3DGSV-Assessor, a reference-flexible, large multimodal framework that scores all three dimensions in one generation pass and fuses a ranking-based reference cue into the frozen regressor through a bounded gate, supporting both reduced-reference (RR) and no-reference (NR) assessment. Extensive experiments demonstrate state-of-the-art correlation with human perception on 3DGS-VBench+ and strong zero-shot generalization to other NVS benchmarks. The dataset and code will be released at https://github.com/YukeXing/3DGSV-Assessor.