Marr-Hildreth algorithm

The Marr-Hildreth algorithm is an edge detection method for digital images, finding the continuous curves where brightness changes strongly and quickly. It first convolves the image with the Laplacian of the Gaussian function, or with the faster difference of Gaussians approximation, and then detects zero crossings in the filtered result to locate the edges. The Laplacian-of-Gaussian operator is sometimes called the Mexican hat wavelet because of its shape when turned upside down. David Marr and Ellen C. Hildreth are among its inventors. It has two main limitations: it produces responses that are not real edges, called false edges, and its localization error can be severe at curved edges, so the Canny edge detector and differential approaches are now preferred.