Improving Similarity Search of Multidimensional Data by Reducing Query Space

To perform the query in a high dimensional query space, a novel filtering strategy is proposed. Projecting the high dimensional data into a low dimensional space and filtering the query space in the projected space, the query space is reduced and shrunk quickly. At the same time, an effective projecting strategy is proposed to enhance the reducibility of low dimensional space. Moreover, a new indexing structure or MS-tree is designed with a new filtering strategy applied to the range query of ML-tree. Experimental results show that reducing query space can improve the indexing performance effectively and reduce the cost for IO and CPU.