Image scale-smoothing, scale-differentiating and binarizing: a new frame for general-scale-edge (GSE) extraction

Traditional edge detectors are almost all based on the frame of Smoothing plus Differentiating plus Binarizing, some of which such as Canny can not provide rich description for edge information as well as can not make exact edge localization. In this paper, a new concept of General-Scale-Edge (GSE) is proposed and defined, and a novel frame for GSE extraction is constructed. Experimental results show that the proposed approach can exactly extract edges at various scales as well as provide rich forms for the description of edge information in images.