Non-rigid point set registration with multiple features

Abstract

We present a new method for non-rigid registration with multiple features in this work. The proposed method is based on an alternating two-step process: correspondence estimation and transformation updating. We first define two vector features for measuring global and local structural differences between two point sets, respectively. We then combine the two features to build a multi-feature based energy function which provides a novel way to estimate correspondences by minimizing global or local feature differences using a linear assignment solution. To enhance the interaction between the two steps, we design an annealing scheme to gradually change the energy minimization from local to global feature differences and the thin plate spline transformation from rigid to non-rigid during the registration process. The registration results demonstrate that our method outperforms four state-of-the-art methods in most experiments.