Incremental Indexing for High-Dimensional Data using Tree Structure

A suitable indexing scheme for multimedia information retrieval is the need of the hour and captured the attention of researchers. In content-based image retrieval applications, users wait for long query time to retrieve similar images from the dynamic growing database. A novel indexing scheme is proposed in this paper to achieve a fast response and higher precision of retrieval. The Dynamically Rearranged High Dimensional Data-tree is constructed using available high-dimensional data with dimensionality reduction based on the occurrences of those values. The high-dimensional data or vectors are represented in the compact form and occupy less memory space. The vectors are clustered based on the patterns to support indexing by avoiding both overlapping of vectors and extra factors for clustering. There is no indication of degradation while database size is changed and it also prevents the effect of noisy vectors during retrieval, which help to improve the retrieval speed. Incremental indexing is an important functionality of the proposed tree, which is suitable for dynamic database. For experimental purpose, the coral image database used and found that the performance of the proposed indexing scheme is encouraging.