An automated method for labelling layers of clustered elements in two dimensional images based on morphological operations is presented here. Unlike the various distance transforms for regular lattices, the procedure deals successfully with clusters composed of different shaped and sized elements and provides a distance measure, in number of layers, from each element in the cluster to any other reference element (or elements). The method was applied to histological sections of two different types of polystratified epithelia. Local clusters in the different cell profile layers can subsequently be used to define the orientation of the tissue layers relative to each element in it.