A Computational Approach to Edge Detection A （一个计算边缘检测方法）.pdf

EE6358 Computer Vision A Computational Approach to Edge Detection A Summarized Report Tanmay Zargar June 7, 2004 © The University of Texas at Arlington Organization of the paper Introduction Mathematical formulation – Detection, Localization and Multiple responses Optimal Detectors by Numerical Optimization Detector for Step Edges An efficient approximation 2 The University of Texas at June 7, 2004 Arlington Organization of the paper (contd.) Noise estimation and thresholding Multi dimensions Need for multiple widths Need for directional operators 3 The University of Texas at June 7, 2004 Arlington One dimensional formulation Identification of the performance criteria. – Detection – Localization – Single response to an edge Mathematical formulation 4 The University of Texas at June 7, 2004 Arlington Mathematical Formulation Detection analogous to SNR. Localization formulated as the reciprocal of the deviation of point marked as edge from the center of true edge. Trivial maximization of the SNR-Localization product. Elimination of maxima in the vicinity of the edge; spatial scaling. The University of Texas at 5 June 7, 2004