Figure 17.164.  Applying example for the “Difference of Gaussians” filter

Original image

Filter “Difference of Gaussians” applied with radius 1 = 1.000 and radius 2 = 0.100.

This filter does edge detection using the so-called “Difference of Gaussians” algorithm, which works by performing two different Gaussian blurs on the image, with a different blurring radius for each, and subtracting them to yield the result. This algorithm is very widely used in artificial vision (maybe in biological vision as well!), and is pretty fast because there are very efficient methods for doing Gaussian blurs. The most important parameters are the blurring radii for the two Gaussian blurs. It is probably easiest to set them using the preview, but it may help to know that increasing the smaller radius tends to give thicker-appearing edges, and decreasing the larger radius tends to increase the “threshold” for recognizing something as an edge. In most cases you will get nicer results if Radius 2 is smaller than Radius 1, but nothing prevents you from reversing them, and in situations where you have a light figure on the dark background, reversing them may actually improve the result.

You can find this filter through Filters → Edge-Detect → Difference of Gaussians….