AI 中文总结
LiTe-GS通过随机子集评估候选视角，将3D高斯泼溅中下一最佳视角选择的oracle复杂度降至O(M log(1/ε))，在保持重建质量的同时大幅减少信息评估次数。
AI 中文摘要
在3D高斯泼溅中，选择信息量丰富的相机视角对于高效训练和自适应细化至关重要，因为每次观测都会显著影响模型参数。然而，基于信息的视角选择策略随着候选视角数量的增加，可能需要反复评估昂贵的信息增益oracle。我们提出LiTe-GS，一种用于3D高斯泼溅中下一最佳视角选择的oracle高效方法。LiTe-GS通过随机子集评估候选视角而非穷举评分整个候选池，减少了信息oracle评估次数。所提方法实现了相对于候选视角数量$M$的期望$O(M\log(1/\epsilon))$ oracle复杂度，且与选择基数$K$无关，同时通过$\epsilon$在oracle效率与近似质量之间提供了显式权衡。我们在所提选择方案下提供了关于oracle复杂度和近似性能的理论保证。在Blender和Mip-NeRF 360上的实验表明，LiTe-GS在保持与基于Fisher信息基线相当的 reconstruction 质量的同时，在不同采集设置下显著减少了Fisher-oracle评估次数。
英文摘要
Selecting informative camera views is critical for efficient training and adaptive refinement in 3D Gaussian Splatting, where each observation significantly influences model parameters. However, information-driven view-selection strategies can require repeated evaluations of expensive information-gain oracles as the number of candidate views increases. We propose LiTe-GS, an oracle-efficient method for next best view selection in 3D Gaussian Splatting. LiTe-GS reduces the number of information-oracle evaluations by performing randomized subset evaluation of candidate views rather than exhaustively scoring the full candidate pool. The resulting approach achieves expected $O(M\log(1/ε))$ oracle complexity with respect to the number of candidate views $M$, independent of the selection cardinality $K$, while providing an explicit trade-off between oracle efficiency and approximation quality through $ε$. We provide theoretical guarantees on oracle complexity and approximation performance under the proposed selection scheme. Experiments on Blender and Mip-NeRF 360 demonstrate that LiTe-GS maintains reconstruction quality comparable to Fisher-information-based baselines while substantially reducing the number of Fisher-oracle evaluations across different acquisition settings.
