Paper Reading

 

Efficient View Clustering and Selection for City-Scale 3D Reconstruction, International Conference on Image Analysis and Processing. Springer, Cham, 2022.

View clustering: 在针对大规模场景的MVS重建中,视角被有重叠地聚类并分别重建和合并为一个场景。

View selection: 已经拍好的照片中选出重要的用于重建。 [ref]

这篇文章采用的Selection中:

optimal means the smallest subset of cameras that guarantees that each point in the cluster is seen by at least $N_{vis}$ cameras and each camera have at least $N_{match}$ other cameras to be successfully matched with.

其中:

Two cameras are considered to be matchable if they see a sufficient number of common points. This differs from previous literature, where the typical similarity measure is the average Gaussian-weighted triangulation angle between the two camera centers and the common keypoints[5, 9]

然后解一个Integer Linear Programming (ILP) 整数线性规划问题。

这个CVPR10 Towards Internet-scale Multi-view Stereo 中给了一个两个图片MVS重建一个点准确度的衡量: