Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
412043 | Neurocomputing | 2015 | 6 Pages |
In recent years, we have witnessed a flourishing of 3D object modelling. Efficient and effective 3D model retrieval algorithms are high desired and attracted intensive research attentions. In this work, we propose a view-based 3D model retrieval algorithm based on weighted locality-constrained group sparse coding. Representative views are first selected by clustering and the corresponding weights are provided by considering the relationship among these views. By grouping the views from 3D models, a locality-constrained group sparse coding method is employed to find the reconstruction residual for each query view. The distance between query model and candidate model is taken as the weighted sum of residual. The query model is matched to the model which can best reconstruct the query model. Experimental comparisons have been conducted on the ETH 3D model dataset, and the results have demonstrated the effectiveness of the proposed method.