In this paper, we propose a new similarity measure to combine multiple views of 3-D objects for non-rigid registration. It is a metric that integrates multiple 2-D view features representing a visual identity of a 3-D object seen from different viewpoints. The robustness to non-rigid distortions is achieved by the proximity correspondence manner. The human face, a typical non-rigid object, was chosen to evaluate the capability of the proposed object matching technique. Very encouraging results were obtained which showed that the proposed Significant-Based Multi-View Hausdorff Distance (SMVHD) provides a new fusion method for non-rigid 3-D object registration.
展开▼
