A study on Nearest-Vector Tree

From the past two decades, the research area of nearest neighbor search in high dimensional data sets has always been in the limelight. Content-based multimedia indexing has been an active area of research as multimedia content is mapped into high-dimensional vectors of numbers, which are then stored in a high-dimensional index. For large collections, high-performance environments and large amount of main memory have been used. This paper reviews the NV-Tree (Nearest Vector Tree), a disk based data structure, which addresses the specific problem of locating the k-nearest neighbors within a collection of high dimensional data sets. The NV-tree is already used in industry to index more than 150 thousand hours of video for (very effective) near-duplicate detection. We present a critical summary of published research literature pertinent to NV-Tree under contemplation for research. The purpose is to create familiarity with existing thinking and research on a particular topic, which may justify future research into a previously overlooked or understudied area.