Speeded up robust features
lastmod 2026-11-10created 2022-07-122 edits
Speeded up robust features (SURF) is a computer vision algorithm that detects and describes scale- and rotation-invariant keypoints in images. It builds on the ideas of SIFT but greatly increases speed by approximating the Hessian matrix with integral images and box filters, and it creates a 64-dimensional descriptor from Haar wavelet responses around each keypoint. It is used for object recognition, image registration, and 3D reconstruction, and due to patent issues it is included in OpenCV's contrib module.