| Abstract: | 3D model reconstruction of defective bone is of great practical value in the application for medical diagnosis of bone, plastic,
prosthesis surgery planning and so on. Based on Delaunay triangulation, an algorithm of surface reconstruction is proposed
to reconstruct surface of defective bone in this paper. In the algorithm, data of CT 2D-images are input to build the space
point dataset. The bounding box theory is applied to search k-nearest neighbors of pi, and its normal vector is computed.
K-nearest neighbor points of pi are projected on the tangent plane, and its Delaunay 2D-neighbors are chosen from those projected
points. Delaunay 3D-neighbors of pi are gotten by mapping Delaunay 2D-neighbors. Delaunay 3D-neighbors are searched and matched
to construct triangular mesh of surface. Then, the topology of Delaunay 3Dneighbors is reconstructed to identify the feature
of hole. Subsequently, a weighted undirected connected graph is generated. EMST of this graph is generated by using Prim algorithm.
After searching EMST based on the depth-first principle, the normal vector is adjusted to orient the same direction of surface.
Lastly, according to the constructed topological relation, 3D geometric model of defective bone is reconstructed. The validity
and reliability of algorithm are verified by computing a typical example, and its time and space complexity is lower. |