SURF is a speeded-up version of the SIFT algorithm that was introduced in 2006. It speeds up the feature detection process compared to SIFT by approximating the Laplacian of Gaussian with box filters, which can be easily calculated using integral images. SURF also uses wavelet responses in horizontal and vertical directions and applies Gaussian weights within a neighborhood to determine keypoint orientations. It provides 128-dimensional feature descriptors by separately summing horizontal and vertical wavelet responses within 4x4 subregions around each keypoint. This makes SURF faster than SIFT while maintaining similar performance, though it is less robust to viewpoint and illumination changes.