An efficient tree structure for indexing feature vectors

Abstract

This paper addresses the problem of feature indexing in feature vector space. A linked-node m-ary tree (LM-tree) structure is presented to quickly produce the queries for an approximate and exact nearest neighbour search. Three main contributions are made, which can be attributed to the proposed LM-tree. First, a new polar-space-based method of data decomposition is presented to construct the LM-tree. Second, a novel pruning rule is proposed to efficiently narrow down the search space. Finally, a bandwidth search method is introduced to explore the nodes of the LM-tree. Extensive experiments were performed to study the behaviour of the proposed indexing algorithm. These experimental results showed that the proposed algorithm provides a significant improvement in the search performance for the tasks of both exact and approximate nearest neighbour (ENN/ANN) searches.