A Skeletonizing Reconfigurable Self-Organizing Model: Validation Through Text Recognition
Neural Processing Letters
Hi-index | 0.00 |
A variety of techniques in machine vision involve representation of objects by using their shape skeleton. In this paper we present a method to obtain the skeletal shape of binary images in the presence of both boundary noise and noise occurring inside object regions. We propose to obtain the skeletal shape of such images by a modified version of the Kohonen self-organizing map, implemented in a batch processing mode. The modifications allow the map to adapt to the input shape distribution. At each iteration, a competitive Hebbian rule is used to progressively compute the Delaunay triangulation of the shape. Information from the triangulation augments the map topology to yield the final skeletal shape. The batch mode implementation of the self-organizing process, allows our approach to compare very favourably, in terms of computational time, with the traditional flowthrough implementations. Encouraging experimental performance has been obtained on a variety of shapes under varying signal to noise ratios.