Time: Monday, February 06, 2017, 1:00pm @ MTH3206

Speaker: Rafael Reisenhofer (University of Bremen)

Title: Edge, Ridge and Blob Detection with Symmetric Molecules

Abstract: We present a novel approach to feature detection, in which the symmetry properties of directionally sensitive analyzing functions constructed in the so-called alpha-molecule framework can be exploited to detect edges, ridges and blobs in images. Furthermore, we show how additional information, such as the widths of ridges and blobs or local tangent orientations of edges and ridges can be extracted by analyzing the behavior of coefficients corresponding to differently scaled and oriented symmetric molecules. For each type of feature, a measure is defined that assigns values ranging from 0 to 1 to each point in an image denoting the likelihood that an edge, ridge or blob is centered at the respective location. These measures are based on inner products of the considered image with differently scaled and oriented alpha-molecules generated by applying rotation and scaling operators to even- and odd-symmetric generating functions. The measures are by construction contrast invariant, stable in the presence of noise and capable of detecting features arising both from smooth and sharp transitions. This is joint work with Emily J. King. Back to seminar