Center-based indexing for nearest neighbors search

We address the problem of indexing data for the k nearest neighbors (k-nn) search. We present a tree-based top-down indexing method that uses an iterative k-means algorithm for tree node splitting and combines three different search pruning criteria from BST, GHT and GNAT into one. The experiments show that the presented indexing tree accelerates the k-nn searching up to several thousands times in case of large data sets.