A robust approach for constructing a graph representation of articulated and tubular-like objects from 3D scattered data

This paper describes an approach for constructing a graph representation of 3D objects and more particularly of articulated and tubular-like objects. For objects without cavities, this representation is a tree structure that encodes the object template while being invariant to global and local rigid transformation. The approach described in this paper has some interesting aspects: (1) It operates on raw 3D scattered data points, without any pre-processing stage. (2) It has low computational cost. (3) It is robust against irregular data point distribution and data deficiencies. This graph representation can be used in various applications such as object coding, recognition, and segmentation.