In this section, it is assumed that a homogeneous set of pixels have been defined which are connected in some way to form a region or regions. A region may be described by features, such as the centroid, the area, the perimeter length etc. or indeed defined by the enclosing boundary, using the types of representation discussed already for boundary segments. Some other common representations are considered which may be used either for data compression or as the constituents of a model for higher order processing.

In general, the number of regions in an image is a variable, in contrast to the number of pixel elements which is fixed according to the image dimensions. Probably the simplest technique for regional representation is run-length coding