3D shape recursive decomposition by Poisson equation

This paper proposes a novel algorithm that decomposes the 3D shape into meaningful parts based on Poisson equation. The whole algorithm is divided into three steps. Firstly, shape signature is defined with Poisson equation. Secondly, the binary decomposition based on shape signature is recursively performed to get a coarse decomposition result. Finally, the graph-based minimum cut is used to refine the jaggy boundaries in the initial result. The proposed algorithm not only obtains a set of meaningful parts, but also is robust in the case of deformation, rotation and other transformations. Furthermore, it can process large 3D shapes in an efficient way.