Canny edge detector (Canny filter) for image processing and computer vision

N. Petkov and M.B. Wieling, University of Groningen,
Department of Computing Science, Intelligent Systems

This page contains examples concerning an simulation program available on internet.

With the applet you can use the Canny filter for edge detection. You can add to the Canny edge detector a surround suppression step that enhances object contours by inhibiting texture edges. You can also visualize the gradient and the derivatives of a Gaussian function that are used to compute the gradient.


0) Quick start

a) Press the button "Update view!" using the following default parameter values: Observe the edges of the squares in the output image window.

b) To trace back how this output has been computed, uncheck now the parameters "Enable thinning" and "Enable hysteresis thresholding of edge strength" and press the button "Update view". Thinning and thresholding are now disabled and you can observe the gradient magnitude in the output window.

c) To visualize only one of the components of the gradient, e.g. the component on the horizontal (x-) axis, select 0 (degrees) in the parameter window for "Displayed orientations" at the bottom of the parameter page and press the button "Update view!". Observe the result in the output image window. This component of the gradient is large in positions where an intensity transition in the horizontal direction is encountered. Positive (bright) and negative (dark) values correspond to different polarities of the transition from dark to bright and bright to dark, respectively.

d) Increase now the value of the parameter sigma, e.g. sigma=3, to observe how the response is broadened.

e) The observed response is computed by convolving the input image with a derivative of Gaussian function. Select "Filterkernel (x)" in the select window above the output image in order to visualize the horizontal (x-) derivative of the Gaussian function.

Further examples will be added in the near future.

Last changed: 2008-07-03