doi:10.1016/j.cad.2007.03.008
Copyright © 2007 Elsevier Ltd All rights reserved.
A divide and conquer algorithm for medial surface calculation of planar polyhedra
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Ian Strouda,
,
, Gábor Rennerb and Paul Xirouchakisa
aEcole Polytechnique Fédérale de Lausanne, DGM-ICAP-LICP, ME-Ecublens, CH-1015 Lausanne, Switzerland
bComputer and Automation Research Institute (SZTAKI), Hungarian Academy of Sciences, Kende utca 13-17, H-1111 Budapest XI, Hungary
Received 9 December 2005;
accepted 27 March 2007.
Available online 13 April 2007.
Abstract
The Medial Axis Transform surface, (or MAT or MS) is proving to be a useful tool for several applications and geometric reasoning tasks. However, calculation of the MAT is a time-consuming task and the benefits of the mathematical-based tool are offset by the cost of the calculation. This paper presents a method for medial surface calculation which uses subdivision to simplify the problem and hence speed up the calculation, a so-called ‘divide-and-conquer’ approach. The basis for this is a modification of the dual structure of the original object. As the calculation proceeds this structure is broken up into sub-pieces each representing a simpler sub-part of the MAT. Perhaps more importantly, this method creates a correct node decomposition of the dual structure. The paper goes on to demonstrate some applications of the results for geometric tasks, specifically offsetting and model subdivision, which are normally expensive but are simpler based on the MAT calculation results.
Keywords: Medial axis transform; MAT
Fig. 1. Illustration of the MAT dual structure and MAT.
Fig. 2. Illustration of convex and concave edges.
Fig. 3. An example Boundary Representation data structure.
Fig. 4. Original object and MAT (from [25]).
Fig. 5. (a) Annotated MAT and (b) Delaunay nodes (from [25]).
Fig. 6. Accretive construction of the Delaunay nodes (from [25]).
Fig. 8. MAT calculation using the divide and conquer process (from [25]).
Fig. 9. MAT calculation process steps (from [21]).
Fig. 10. Corner block and MAT.
Fig. 11. Divide and conquer dual structures and Delaunay nodes from subdivision.
Fig. 12. Vertex with two concave edges.
(a) Original pocket.
(b) Normal dual of v.
(c) Concave edges split.
(d) modified dual of v.
Fig. 13. Modified dual around concave vertex.
Fig. 14. Subdividing faces with one concave edge node.
Fig. 15. Square torus, MAT and Delaunay nodes (from [25]).
Fig. 16. Separation at necks.
Fig. 17. Arrangement of nodes in topological dual.
Fig. 18. Adjacent Delaunay node faces to be extracted.
Fig. 19. Extracted faces.
Fig. 20. U-shaped block, MAT and node structure.
Fig. 22. Offset 0.5, all nodes joined.
Fig. 23. Offset 1, nodes 4, 5, 6 and 7 disappear.
Fig. 24. Offset 1.5, same topology as Fig. 23.
Fig. 25. Offset 2, only nodes 10, 11, 12 and 13 remain.
Fig. 26. All offset objects.
Fig. 27. Joined node substructures.
Fig. A.1. Cube structure elements.
Fig. A.2. Original edge and dualled edge.
Fig. A.4. Vertices of the dual.
Fig. A.5. Winged-edge structure pointers.
Fig. A.6. Addition of first edge in dual structure.
Fig. A.7. Addition of second edge in dual structure.
Fig. A.8. Addition of third edge in dual structure.
Fig. A.9. Final dual structure.
Fig. A.10. Concave edges in dual structure split.
Fig. A.11. Concave edge adjacent faces in dual structure split.
Fig. A.12. Concave vertex face in dual structure split.
Table A.1
Original and dualled body elements

Table A.2
Structure lists for corner block


Corresponding author. Tel.: +41 21 693 2949; fax: +41 21 693 3553.