Gradient magnitude (which can be thought of as the strength of edges) computed using (forward) finite-difference approximations of the partial derivatives can be used for edge detection, as we saw earlier. The following screenshot shows the output obtained by using the same code as the previous time to compute the gradient magnitude, and then clip the pixel values in a [0, 1] interval, with the zebra's input gray-scale image:
The following screenshot shows the gradient magnitude image. As can be seen, the edges appear to be thicker and multipixel wide:
In order to obtain a binary ...
