Decomposition of grey-tone patterns using ridge-information

We present a two-stage procedure to decompose a grey-tone pattern into a number of parts, each one found in correspondence with perceptually meaningful higher intensity regions. The fìrst stage is somebow related to grey-skeletonization, since it gives evidence to feature sets constituted by pixels placed along ridges and midlines of protruding regions of the pattern (interpreted as a Digital EIevation Model). The second stage establishes the spatiaJ relations amoog the feature sets and originates the parts. These are obtained by finding, and possibly merging, the zones of influeoce of the feature sets detected on the distance transform of the grey-tone pattern. Since a set of alternative pattern organizations into parts can be obtained, a criterion for the selection of the most significant one is proposed.