
248 A Computational Introduction to Digital Image Processing, Second Edition
(a) Zero crossings (b) Using an LoG filter first
FIGURE 9.27: Edge detection using zero crossings
Note that the image must be converted to type
float64
first. The filtering function applied
to an image of type
uint8
will return an output of the same type, with modular “wrap
around” of values outside the range 0–255. This will introduce unnecessary artifacts in the
output.
9.9 The Canny Edge Detector
All the edge finding methods so far have required a straightforward application of linear
filters, with the addition of finding zero crossings. All of our systems support a more
complex edge ...