Automated threshold driven edge detection using 2-way difference calculation and neighborhood comparison

An edge in an image is a point where discontinuities exist, or where the brightness changes considerably. The paper puts forward a method that detects the presence of edges by evaluating the difference in intensity levels between adjacent pixels on either side of a central pixel both horizontally and vertically throughout the image. The novelty of this edge detection procedure lies in the automation of the threshold generation based on the original image. This is followed by reconstruction of undetected edges by means of comparison with the original image. The paper then provides the output after a stage of skeletonization of the reconstructed image.