Improved metrics in image processing applied to the Hilditch skeleton

An algorithm is presented in which binary object pixels are addressed and processed in order of increasing distance from the background. The distances are defined as path lengths. The metric can be chosen to obtain arbitrarily good approximations of the Euclidean metric. The algorithm incorporates an efficient propagation method in which extensive use is made of directional information. It is applied to the Hilditch skeleton and proves to be as fast as the best of the other Hilditch skeletonization algorithms available in software.<<ETX>>