Ultra-Fast Skeleton Based on an Isotropic Fully Parallel Algorithm

Antoine Manzanera (1, 2, 3) , Thierry Bernard (2) , Françoise Prêteux (1) , Bernard Longuet (3)

1 TSP - ARTEMIS - Département Advanced Research And Techniques For Multidimensional Imaging Systems
2 DGA/CTA/DT/GIP - CEP Arcueil
3 Aérospatiale-Missiles

In this paper we introduce a new thinning algorithm, called MB, which is optimized with respect to the total number of elementary Boolean operators needed to perform it. We first emphasize the sound foundations of the algorithm, which is built by expressing into the Boolean language the three following constraints: (1) homotopy, (2) median axis and (3) isotropy. The MB algorithm benefits from both novel algorithmic ideas and systematic logic minimization. By hunting down any redundancy in the expressions of topological/geometrical features , we achieve a procedure that is: dramatically low-cost, as it is completely computed in 18 Boolean binary operators per iteration, and secondly, fully parallel, or one-single-pass, which guarantees that the number of iterations equals half the biggest object thickness.

Mots clés: Thinning; digital skeleton; discrete topology

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Traitement des images [eess.IV]; Géométrie algorithmique [cs.CG]

Antoine Manzanera, Thierry Bernard, Françoise Prêteux, Bernard Longuet. Ultra-Fast Skeleton Based on an Isotropic Fully Parallel Algorithm. Discrete Geometry for Computer Imagery (DGCI'99), Mar 1999, Marne-la-vallée, France. ⟨10.1007/3-540-49126-0_24⟩. ⟨hal-01245393⟩