3D Model Retrieval Algorithm Based on DSP-SIFT Descriptor and Codebook Combination

Recently, extensive research efforts have been dedicated to view-based 3D object retrieval, owing its advantage of using a set 2D images represent objects. Some existing image processing technologies can be employed. In this paper, we adopt Bag-of-Words for retrieval. Instead SIFT, DSP-SIFT is extracted from all as features. Moreover, two codebooks the same size are generated by approximate k-means. Then, combine correct quantization artifacts and improve recall. Bayes merging applied address codebook correlation (overlapping among different vocabularies) provide benefit high Approximate Nearest Neighbor (ANN) used quantization. Experimental results on ETH-80 datasets show that our method improves performance significantly compared with state-of-the-art approaches.