Remaining pairs:

Considering n pairwise feature correspondences we want sets of cor-responding matching features across multiple images, i.e., tracks. Track identification in a set of images is an important task in computer vision. It allows solving geometry-related problems like: • region tracking, match-moving, video stabilization, • image-stitching, • structure from motion [4] and odometry. We present an efficient algorithm to fuse two-view correspondences into multi-view consistent tracks that is: • simple, • efficient: lower computational complexity than [2, 3], • reliable: able to identify all tracks, contrary to [2, 3]. THE TRACK COMPUTATION PROBLEM The problem of feature point tracking consists in linking multi-view correspondences from pairwise matches that share a common point. It consists of a correspondence fusion problem