An efficient adaptive algorithm for edge detection based on the likelihood ratio test
Alberto De Santis, Daniela Iacoviello · International Journal of Adaptive Control and Signal Processing · 2002
Abstract The edge detection problem in blurred and noisy 2‐D signals is dealt with. An adaptive signal processing algorithm is proposed which marks edge points according to an hypothesis test which compares the likelihoods of two models describing the local signal behaviour in the two cases of absence/presence of an edge. The two models are identified by a regularized least squares estimation algorithm, obtaining a numerically efficient procedure, quite robust with respect to additive noise and blurr perturbation. No global thresholding or data prefiltering is required. Copyright © 2002 John Wiley & Sons, Ltd.