Robust and effective shape prior modeling from a set of training data remains a challenging task, since the shape variation is complicated, and shape models should preserve local details as well as handle shape noises. To address these challenges, a ...