Optimal ramp edge detection by orthogonal wavelet transform

Abstract

A method for multiresolution edge detection is described which is based on wavelet representation in which the edges are detected from the local maxima of the detail images in the orthogonal wavelet transform. The new idea is to compute an optimal edge detection filter for ramp edges. It is shown that the approximation of an edge at each scale in its orthogonal wavelet decomposition can be modeled by a ramp contour whose equation is given. Good edge detection results are obtained. Such a multiscale edge detector is important not only for the usual applications such as pattern recognition and image segmentation, but also for object-based image coding where the reconstruction of the original image from the edges at different scales is made possible by wavelet analysis-synthesis. >