A novel shape descriptor based on empty morphological skeleton subsets

In this paper, a novel shape descriptor, based on empty morphological skeleton subsets, is proposed. This new descriptor possesses many properties required for practical shape recognition i.e. it is non-specific, robust to scale, rotation and boundary variations. One advantage of the proposed descriptor is that the empty morphological skeleton subset pattern may be used to distinguish different objects even with an identical union of skeleton subsets. Implementation of this descriptor, for recognition of objects, has been successfully realized using neural networks.