Curve-Skeleton Extraction from 3D Objects Using Spiral Patch

Abstract

The extraction of curve-skeleton from a 3D object is an essential process for different computer vision applications such as 3D objects matching and recognition, 3D objects segmentation, 3D animation, and 3D surface reconstruction. In this paper, we propose a robust and efficient method for skeleton extraction from general 3D objects based on weight graph representation. We define first a novel spiral patch for graph vertices as the shortest loop starting from a given vertex over the surface contour. We construct then a curve skeleton graph CSG using spiral patches that will represent the skeleton. We demonstrate the robustness of this method through different experimental results with respect to noise and different transformations. Moreover, we present two applications based on our method, which are shape partitioning, and gesture recognition