A Study on Edge Location Error of Second Derivative Edge Detection Operators
二阶导数型边缘检测算子边缘定位误差的研究

Two edge detection operators of second derivative model, the discrete Laplacian of Gaussian (DLOG) and Laplacian of binomial distribution (LOB) operators are introduced in this paper. A digital edge contaminated by an independent Gaussian noise is used as an edge image model. After convolving the image model with the edge detection operators of second derivative model, a plane fitting in the least squares sense is used to detect the zero crossing points as edge points. The Probabilities P_d of edge location and the Probabilities P_f of false edge detection are derived to compare the performances of various edge detection operators of second derivative model.