Multi-scale skeletons via permanence ranking

Multiresolution shape representations are highly desirable, as they provide a flexible tool that fits the user's needs hetter than single resolution systems. In this paper, the skeletons of a 2D pattern recorded onto a binary AND-pyramid are extracted at all resolution levels. The so obtained skeletons are then transformed into muItiresolution skeletons by ranking skeleton subsets, based on their permanence in the skeleton at the various scales. The potential of multiresolution representation is exploited in this way, since the skelcton is available at various scales and, at each scale, skeleton components are furtherrnore ranked according to their perrnanence along the pyramid levels.