Video Synchronization (2003-2004)
When you have two videos of a moving 3D scene, you can compute all sorts of useful information from the images but only if you know which pairs of images (one from each camera) were captured at the same time. Otherwise, you end up effectively comparing images of two different scenes which gives you garbage.
This work looks at how two videos of a human being doing something dynamic can be synchronized semi-automatically by looking for pairs of frames that might possibly correspond to the same 3D scene seen from different angles.