Enhanced 3D shape retrieval using view-based silhouette representation

Recently, the development of 3D model database systems and retrieval components are becoming increasingly important due to a rapidly growing amount of available 3D models. This has made the retrieval for specific 3D models become a vital issue. Unfortunately, traditional keyword searching techniques are not always effective for 3D object retrieval. In this paper, an enhanced feature, called silhouette descriptor, is proposed for 3D models retrieval. The propose retrieval method is based on a multi-view approach, which keeps 3D models coherence by considering a set of images in specific view direction simultaneously. Experimental results show that the proposed method is faster, simpler and most efficient than other method.