3D mesh model retrieval based on block

潘翔,张引,张三元,叶修梓

DOI: https://doi.org/10.3785/j.issn.1008-973X.2004.12.009

2004-01-01

Abstract:To precisely compute the similarity between different models for content-based 3D retrieval, a 3D mesh model retrieval algorithm based on block similarity was proposed. In the preprocessing stage, an adaptive linear subdivision was performed to make the vertex distribution even over the given mesh, and pose estimation was undertaken to normalize the orientation of the model. The preprocessed model was segmented into a set of blocks, and the normal histogram was constructed for each block. The similarity calculation between 3D models was performed using a local-to-global strategy: The local similarity between blocks was computed, then an improved similarity function based on local block similarity and point density was defined to calculate the global similarity. Experimental results showed that the proposed algorithm could achieve better retrieving precision than other similar retrieving algorithms.