Algorithmic Research On2D Shape Retrieval

The V-system is a special orthogonal multiwavelet, due to the virtue of its orthogonality and multi-resolution, only a small number of basis functions are required to describe characteristics of a signal. And multi level features of the signal, from any coarse scale to a finer scale, can be obtained by increasing the number of basis functions.In this paper, the theory bases on V-system:According to the characteristic of V-system, this paper improves the V-moment definitions and theories further and apply them to image retrieval. The results of experiments conducted on general database of MPEG-7-shape-CE2and self-build database which contained10750trademark images show that the V-moment outperforms some classical moment; This paper applies the characteristic of the V-system on shape similarity retrieval and a new shape retrieval algorithm, V-algorithm, is achieved by combing its boundary features and region features. Firstly, the boundary of the image is extracted and orthogonally decomposed into a V-series, thereby obtaining a boundary feature vector consisting of the V-descriptors. Meanwhile, a region feature vector of the image is obtained by applying the V-transform to the image region. Then similarity between shapes is measured using an integrated shape descriptor, which combines the boundary features and region features. The results of experiments conducted on four benchmark database show that the method in this work has advantage in retrieval efficiency in comparison with several classical algorithms.