1. 摘要
本文展示一套MVS系统，该系统利用非结构化的图片实现鲁棒且稠密的建模。本文的主要贡献是深度和法向量的联合估计，用光度和几何先验进行像素筛选，多视图几何一致项，该项同时进行精修和基于图片的深度和法向量的融合。在标准数据和大尺度网络图片上的实验证明了其在精度、完善性、效率方面的优异性能。
2. 引言
主要贡献
- Pixelwise normal estimation embedded into an improved PatchMatch sampling scheme.
- Pixelwise view selection using triangulation angle, incident angle, and image resolution-based geometric priors.
- Integration of a \temporal" view selection smoothness term.
- Adaptive window support through bilateral photometric consistency for improved occlusion boundary behavior.
- Introduction of a multi-view geometric consistency term for simultaneous depth/normal estimation and image-based fusion.
- Reliable depth/normal filtering and fusion.
2. 代码地址
github.com/colmap/colmap
参考文献
COLMAP： Pixelwise View Selection for Unstructured Multi-View Stereo - 知乎
Pixelwise View Selection for Unstructured Multi-View Stereo
