SURF

SURF (Speeded Up Robust Features) is a high-performance feature detector and descriptor, leveraging integral images for real-time, scale- and rotation-invariant object recognition.

SURF, or Speeded Up Robust Features, is a proprietary local feature algorithm (Bay, Tuytelaars, and Van Gool: ECCV 2006) designed for rapid image comparison and representation. It achieves significant speed gains—several times faster than its predecessor, SIFT—by utilizing an integer approximation of the determinant of the Hessian matrix for keypoint detection. The core optimization relies on integral images, which enable near-constant-time computation of box filters, regardless of size. This efficiency makes SURF a robust choice for real-time computer vision applications, including object tracking, image registration, and 3D scene reconstruction.

https://www.vision.ee.ethz.ch/~surf/eccv06.pdf