抄録
A curve skeleton of 3D complex shape is a compact and simple graph, which provides sufficient information on geometry and topology of the shape. It is very useful for many computer graphics applications involving shape analysis. Much research has been focused on volumetric and polygonal mesh models. However, only a few have been paid attention to curve skeleton extraction from point clouds; particularly from noisy and incomplete data. We propose a robust algorithm for extracting curve skeleton from point cloud. The process starts from estimating centers of antipodes of each point, so called skeletal candidates. Those candidates are absolutely inside the shape; but scattered. We thus filter and shrink them to create less noisy skeletal candidates before applying one-dimensional Moving Least Squares to build a line-like point cloud. We then downsample the computed thin cloud to sparse skeletal nodes. We finally utilize least squares ellipse fitting to relocate skeletal nodes to guarantee that they are under centeredness property of a curve skeleton. We demonstrate the consistency of our method on several models ranging from clean to noisy and incomplete data.
