Apparatus is disclosed for rapidly computing the Gaussian of Laplacian convolution of the point values in an image. Speed and hardware efficiency are achieved through the use of the separability of higher dimensional Gaussians, the binomial approximation to the one-dimensional Gaussian, the zero mean of Laplacian filtered signals, the reduced sampling requirements for low pass filtered signals, and a binary correlator that does many correlations in parallel, allows correction for geometric distortion and variable sampling density.
展开▼
