Area of image processing which deals in separating an image's regions. For example, a microscope image of a group of cells could need segmentation to count the number of visible cells, average size or similar measures.

Segmentation is usually carried out detecting contours, using different approaches. If we treat an image like a height field, then the derivatives would give us an approximation of contours (a high derivative means a high slope, which is characteristic in a contour). Further refinements include the gradient and the laplacian.

Morphological image processing also lets us find contours of the image, particularly by approximating the gradient of the image (for example, substracting the erosion from the dilatation), or by using watersheds.

Another common technique is blurring the image (convolutioning with a gaussian, for example) before segmentation, as many annoying quirks in the image can be removed that way.