doiSerbia

3D mesh skeleton extraction using prominent segmentation

Sun Xiaopeng (School of Mechanical & Engineering, Dalian University of Technology, Dalian, China)
Pan J. (Department of Computer & Information Technology, Liaoning Normal University, Dalian, China)
Wei Xiaopeng (Key Laboratory of Advanced Design and Intelligent Computing (Dalian University), Ministry of Education, Dalian, China)

Skeleton of 3D mesh is a fundamental shape feature, and is useful for shape description and other many applications in 3D Digital Geometry Processing. This paper presents a novel skeleton extraction algorithm based on feature point and core extraction by the Multidimensional scaling (MDS) transformation. The algorithm first straights the folded prominent branch up, as well as the prominent shape feature points of mesh are computed, a meaningful segmentation is applied under the direction of feature points. The Node-ring of all segmented components is defined by discrete geodesic path on mesh surface, and then the skeleton of every segmented component is defined as the link of the Node-ring's center. As to the core component without prominent feature points, principal curve is used to fit its skeleton. Our algorithm is simple, and invariant both to the pose of the mesh and to the different proportions of model's components.

Keywords: Skeleton, MDS, Discrete geodesic path, Node-ring