A thinning algorithm based on prominence detection

摘要

Binary pictures are taken into account and an algorithm which transforms a digital figure into a set of 8-simple digital arcs and curves, by employing local sequential operations, il illustrated. The adopted procedure considers the removal of suitable contour elements of the given figure and it is repeated on every current set of 1-elements as long as the final thinned figure is obtained. The notion of local elongation is employed to find, at every step of the process, those contour regions which can be regarded as significant protrusions. Then such regions are detected before applying the removal operations, contributing hence to the isotropic behavior of the proposed thinning transformation. Further refinements of the algorithm are also discussed.