Optimal detection of blurred edges
J.A. Robinson · 2003
I develop a variational model for the detection of blurred edges. Though yielding a complicated general solution, the model gives insight into edge detection models previously proposed for step edges. In particular it shows that a previous IIR solution for step edges is correct even in the FIR case, and suggests why operators approximating the first derivative of a gaussian are successful, despite their suboptimality for step edge detection.