structure from motion (SfM)

Recovering both a scene's 3D structure and every camera's pose from many overlapping images with no camera positions known in advance.

Structure from motion (SfM) jointly recovers a scene's 3D point cloud and every contributing camera's pose from a set of overlapping images with no camera positions known beforehand, by chaining pairwise correspondence and pose estimates across many image pairs and refining the whole reconstruction (commonly with bundle adjustment) to minimize reprojection error.

Examples

A tourist's 200 unordered photos of a cathedral, with no GPS or camera-orientation metadata, are enough for SfM to recover both a 3D point cloud of the building and every photo's own camera pose.