An Approach of Hierarchical Image Index Based on Subspace Cluster
一种基于子空间聚类的图像分层索引方法

Nowadays large volumes of data with high dimensionality are being generated in many fields. Many approaches have been proposed to index high-dimensional datasets for efficient querying. ClusterTree is a new indexing approach representing clusters generated by any existing clustering approach. Lots of clustering algorithms have been developed, and in most of them some parameters should be determined manually. The authors propose a new subspace-cluster indexing algorithm, which based on the improved CLIQUE and avoids bias on any parameters caused by user. Using multi-resolution property of wavelet transforms to reprocess the distribution curve of samples, the proposed approach can cluster at different resolution and remain the relation between these clusters to construct hierarchical index. The results of the experiment confirm that the subspace-cluster algorithm is very applicable and efficient, and show that this hierarchical indexing structure does well in the content-based image retrieval.