Skip to content
Free public beta until December 4, 2026: every course is open, no card needed. After that, the first chapter of each course stays free and full access is a one-time $39 ($19 for up to 100 founding members chosen for helpful feedback). Your progress is kept. How the beta works

correspondence

A matched pair of points -- one from each of two images -- believed to show the same physical point.

A correspondence is a pair (\mathbf{p}_A, \mathbf{p}_B) of points, one detected in image A and one in image B, proposed to be projections of the same 3D point or the same physical feature. Descriptor matching (Chapter 5) proposes correspondences by comparing feature descriptors; RANSAC then separates the correct ones (inliers) from the wrong ones (outliers) before any downstream geometry -- a homography (Chapter 6), a camera calibration (Chapter 7), or a stereo reconstruction (Chapters 8-9) -- is allowed to trust them.

Examples

  • A rooftop corner visible in both photos of a stereo pair gives one correspondence: its pixel location in the left image paired with its pixel location in the right image.

See also

RANSAC
Read the full section →
Graph settings

Calculator