Camber Descriptor for Content-based 3D Model Retrieval

Hong Tu,Guohua Geng

2013-01-01

Abstract:Along with advancements in 3D data acquisition technology and computer processing, 3D model retrieval has been one of the most important research fields in multimedia information retrieval. Feature extraction is one of the most critical technologies in content-based 3D model retrieval system. We defined a new local feature descriptor camber that is the complex of Gaussian curvature and mean curvature. As a correction to mean curvature and Gaussian curvature, it can overcome the shortcoming of Gaussian curvature and mean curvature without additional computation. The performance of the proposed feature descriptor is compared to another by precision/recall tests. Generally, result shows that the new feature descriptor introduces improvement in the 3D model retrieval system. When matching global feature descriptor, it will get a better result in the retrieval process.