Intensity driven detection of parts in grey-tone patterns

A decomposition method, driven by the intensity infonnation characterizing the pattem subsets having locally higher gray-value, is proposed as a possible tool to investigate the shape of a gray-tcne pattern. Connected subsets of the gray-skeleton of the pattern are regarded as the seeds from which the decomposition of the most perceptually significant region of the pattern originates. The zones of infiuence of the seeds play a central role in this process. Since the zones of infiuence may merge under different constraints, alternative decompositions become available and a persistence criterion is adopted to find the most significant pattern organization into parts.