Abstract: In computational vision, edge detection is an important step in visual information processing to represent the image. Along with rationalizing DOG (difference of Gaussians) function as a suitable edge detection operator in terms of localization in spacial/frequency domains, the regularization theory and the informational completeness, we proposed an electronic network model of edge detection. By adjusting network parameters, the system could detect zero-crossings of an image filtered through the △2Gλ, the regularizational or psychophysical operators as DOG functions. The output of the network was a zero-crossing spectrum with time as the scale space dimension. The prospect of hardware network design is also discussed.