Speeded Up Robust Features (SURF) is a patented algorithm similar to and inspired by SIFT (refer to the Applying Scale-Invariant Feature Transform recipe). SURF was introduced in 2006 and uses Haar wavelets (refer to the Applying the discrete wavelet transform recipe). The greatest advantage of SURF is that it is faster than SIFT.
Take a look at the following equations:
The algorithm steps are as follows:
Transform the image if necessary to get the grayscale equivalent.
Calculate the integral image at different scales, which is the sum of the pixels above and to the left of a pixel, as shown in equation (11.1). The integral image replaces the Gaussian filter in SIFT.
Define the Hessian matrix (11.2) containing second-order derivatives of the grayscale image as function of pixel location p and scale σ (11.3).
Determinants are values related to square matrices. The determinant of the Hessian matrix corresponds to a local change in a point. Select points with the largest...