Speeded Up Robust Features Algorithm

Speeded up robust features, known as SURF, is an algorithm for detecting and describing distinctive keypoints in an image, built as a faster alternative to the scale-invariant feature transform for tasks such as object recognition, image matching and three-dimensional reconstruction. Rather than repeatedly blurring and rescaling the image the way SIFT does, SURF approximates the same underlying blob-detection mathematics, the determinant of the Hessian matrix, using box filters of different sizes applied directly to the original image, and it computes each of these box filters in constant time with the help of a precomputed integral image, regardless of the filter's size. Herbert Bay, Tinne Tuytelaars and Luc Van Gool published SURF in 2006, reporting that their standard implementation runs several times faster than SIFT while remaining comparably robust to changes in scale, rotation and lighting.