Global Biharmonic Framework for Robust Non-rigid Model Retrieval

With the emergence of the SHREC Contest, the retrieval performance of non-rigid 3D shapes has witnessed a great improvement. This paper mainly discusses how to improve the stability and discriminate ability for non-rigid 3D shape retrieval. We first introduce a robust global biharmonic (GBI) feature, which has achieved a balance between global and local shape properties. Then, a combination strategy of GBI and ShapeDNA feature (named GBI-DNACE) is adopted to improve the retrieval performance. The experiment results have demonstrated that the proposed GBI-DNACE feature is superior to state-of-the-art, including SD-GDM, mesh SIFT, MDS-CM-BOF and ShapeDNA.