Journal of Pattern Recognition Research

The recent technological progress in acquisition, modelling and processing of 3D data leads to the proliferation of a large number of 3D objects databases. Consequently, the techniques used for content based 3D retrieval has become necessary. In this paper, we introduce a solution to the problem of 3D objects recognition and retrieval by using shape representation of 3D objects views. We propose a 3D indexing and search approach based on the similarity between views. For the task of  2D views matching, we propose a first shape similarity comparison, based on the correspondence of visual contour parts. These parts are obtained by applying shape segmentation with the Curvature Scale Space (CSS) contour based descriptor, in order to solve scale and invariance problems. We propose to combine this partial search description with a second global region based description by Zernike moments. The results obtained shown the robustness and the efficiency of the proposed approach compared to the mutil-view method using only CSS descriptor, and also compared to three other 3D indexing approaches based on geometrical and shape features.