An important step in image processing is the identification of objects, which are bounded by 1dimensional closed curves in the image plane. A preliminary step is the detection of edges. For this task several algorithms have been proposed. Extremal gradient size are most valuable. A problem is that the gradient does not fit into the pixel map, but instead into the gaps between the pixels. So, 2nd order differential operators like the Laplacian may be preferable. Two adjacent areas of different luminance will yield a pair of chains with positive and negative values on both sides of the border. Preferably, the representation of the results should not only show the values, but also the neighborhood relations of the extrema as a 1dimensional graph.