A new stroke segmentation technique is reported. It is based ontwo essential operations applied to a given character: grouping adjacentsegments into blade-like objects according to change of width asmeasured from global directions, and transforming those objectsaccording to width variation as measured from intrinsic orientations.This technique proves to be effective in dividing Chinese charactersinto overlapping and non-overlapping components. Strokes are thenrestored, with the help of templates, by linking certain non-overlappingcomponents terminating at the same overlapping areas. The resultingframework also lays down a basis for the structural matching of Chinesecharacters
展开▼