A novel efficient 3D object retrieval method based on representative slices

In the last few years, the request for a content-based 3D object retrieval system has become a significant issue. At this point, the principal challenge is the mapping of the 3D objects into compact representations referred to as descriptors, which serve as search keys over the retrieval process. In this paper, a new approach will be proposed for 3D objects indexing and retrieval. The main idea is to normalize the 3D objects to insure invariance with respect to affine transformations, and then characterize them by a set of representative slices (RS) along their three principal axes, transforming the shape-matching problem between 3D objects into similarity measuring between their representative slices. In order to reduce the time required to search without diminishing the relevance of the results, we choose among the extracted slices from the 3D object the ones that give the best representation. To achieve this task, we use the k-means clustering method to pull out the representative slices. For the presentation of the effectiveness and superiority of our approach we conduct a comparison of our approach against 3D Zernike descriptor on 146 3D objects from Princeton Shape Benchmark (PSB) database. Experiment results show that our proposed method is superior to 3D Zernike descriptor.