THE RECURSIVE IMPLEMENTATION OF THE OPTIMAL FILTER FOR THE DETECTION OF ROOF EDGES AND THIN LINES

We give the recursive implementation for the optimal filter (according to Canny's criteria) for the detection of image edges the profile of which can be modelled by an exponential function and which can be assumed to be immersed in white Gaussian noise. We show that such an implementation can lead to considerable computational savings when the edges are roof-like.