Multi-resolution skeletons without explicit image smoothing

Three methods described in this paper deal with the problem of multi-resolution skeletonization without explicit image smoothing. The methods are: the removal of `tips´ during the construction of a skeleton or after a skeleton has been obtained, the removal of skeleton branches during the first few iterations or once only at a predefined iteration; and removal of all skeleton points from the nearest node to the remaining image at a predetermined iteration. Comparison with some smoothing methods is also presented