corner detector

An algorithm that flags image locations whose local gradient structure looks like a corner.

A corner detector (Harris, Shi-Tomasi, or FAST) locates keypoints by examining a small window's local intensity structure: Harris and Shi-Tomasi build a structure tensor from the window's gradient products and classify it by that tensor's eigenvalues, while FAST instead runs a cheap ring of pixel intensity comparisons around a candidate point -- three different routes to the same repeatable, well-localized keypoint.

Examples

Harris flags a chessboard's black-white corners as keypoints but ignores the flat interior of each square, where the local gradient structure gives no reliable position to lock onto.