在《OpenMVG: Open Multiple View Geometry》这篇论文中,提到了MVG与MVS的区别和联系:
MVG用于重建相机位置和旋转朝向,同时可以得到稀疏的3D点云。
MVS用于稠密的3D重建。输出形式可以是稠密点云,网格,一系列平面等等。
MVS依赖于MVG的算法结果。
论文原文:
We make here a distinction between multiple view geometry (MVG) and multiple view stereovision (MVS) software. The former is concerned with recovering camera locations and orientations from the data (images and camera intrinsics); it delivers also a sparse set of 3D points, built by triangulation from the feature points observed in the photographs. The latter deals with the dense 3D reconstruction; its output can be a dense point cloud, a faceted surface (mesh), or a set of planes, which can be visualized as a realistic 3D rendering of the scene. It relies on MVG to achieve that.