A hybrid shape decomposition using hyperquadrics and mathematical morphology

This paper discusses a hybrid shape decomposition method using hyperquadrics to describe a part, mathematical morphology to find an initial ellipse, and polygonal approximation to find additional terms of the hyperquadrics. We adopt a weighted least square fitting to find hyperquadrics using the Levenberg-Marquardt optimization method. The proposed method is illustrated in decomposing some 2D binary images such as a natural object and a man-made object. It can be easily extended to 3D range data. The experimental results show that our method is a promising approach in shape representation