Contour Pixel Classification for Character Skeletonization

In this paper it is proposed a mechanism for implementing the isotropic propagation of the figure border to obtain the skeleton of elongated shapes. The mechanism allows for detecting, classifying and labelling the contour pixels depending on the characteristics of the wavefronts which interact during the propagation. The skeleton provided by the algorithm is not affected by the distortions which arise in correspondence of regions where the parts of the figure interact. Moreover, it is given in terms of a set of digital lines, each one corresponding to one of the figure parts, rather than by a connected set of pixels.