Edge detection is one of the most important subjects in image processing. It finds wide applications in pattern recognition, scene analysis, and 3-D vision, because the edges correspond in general to the important changes of physical or geometrical properties of objects in the scene, and they are widely used as primitives in pattern recognition, image matching, etc.

The edges coincide, generally speaking, with gray level transition, so they can be detected by maxima of gradient or the zero-crossing of the second derivatives calculated by some differential operators. Because the differential operators are sensitive to noise, a preprocessing such as smoothing is often necessary to eliminate the noise.

There are three types of edge detection:

• Edge detection using the maxima of gradient, i.e., maxima of the first order derivative.
• Edge detection using the zero-crossings of second order derivative along the gradient.
• Edge detection using an optimal difference recursive filter.

Can be applied only to 8bit grayscale images with one channel.

See EdgeDetection Method for more information.