Text this: Image skeletonization based on combination of one- and two-sub-iterations models