3D model retrieval based on multi-view SIFT feature

Abstract

We research a 3D model retrieval algorithm on the basis of multi-view SIFT features in this paper. By projecting the 3D model from multiple viewpoints, omnidirectional 2D depth images are obtained and from which SIFT features are extracted. Based on k-means clustering algorithm, we establish the codebook according to the proportion of SIFT features number of various shape types and whole SIFT features respectively. As a result, the former is much faster, so it is utilized to build the codebook in our paper. All the SIFT features associated with a model are clustered to generate a simplified vector by a histogram manner. For the similarity matching, Kullback-Leibler divergence is used to calculate distance between simplified vectors. Experiments show the algorithm based on multi-view SIFT feature can gain a satisfactory retrieval result for multiple shape benchmarks. © 2011 Springer-Verlag Berlin Heidelberg.