The canny edge detection filter takes a field with one component (representing a grayscale image) and attempts to detect the edges of objects within the image. There are four steps in the algorithm and several subfilters are used internally. It is not important to fully understand how all the subfilters work to use the edge detection filter but it is helpful to know a little about the first and last subfilter so that you can adjust the four parameters to improve your edge detection results.
The variance and maximum error parameters control the smoothing subfilter that is applied at the start of the canny alogrithm.
The variance governs the size of the convolution kernel used to smooth the image. (See the help on the Discrete gaussian filter for more information on convolution kernels). Increasing the variance means a larger kernel is used. This has the effect of smoothing the image more, reducing the level of image noise at the expense of some detail. Reducing the maximum error level also helps in removing noise. Removing noise is important as noise can result in the detection of undesirable edges.
The upper threshold and lower threshold parameters control the final subfilter which uses a form of thresholding to determine start and end points for edges.
The upper threshold sets the level which a point must be above to use it as the start of the edge. The edge will then grow from that point until the level drops below the lower threshold. Increasing the upper threshold will decrease the number of edges detected. increasing the lower threshold will reduce the length of edges.
Click the "Apply filter" button to update the canny edge detection filter to use the appropriate values.