HnSRTree : the SR-tree library
This is an implementation of the SR-tree [KS97]. The SR-tree is a
multidimensional index structure designed for the acceleration of the
nearest-neighbor search of high-dimensional points in the Euclidean
space. The SR-tree is a disk-based index structure and applicable to
large data sets, say, a database containing more than 10,000 points.
The efficiency of the SR-tree depends on multiple factors, i.e., data
distribution, dimensionality of points, etc. According to our
experimental evaluation, the SR-tree provides satisfactory performance
for the data sets containing more than 40,000 points in 24 to 48
dimensional space which are feature vectors (color histograms) of
natural photo images.
The major part of this library is written in C++. However, the
interface for the C language programs is also provided.
Table of contents
Any feedback is appreciated (corrections, suggestions, etc.).
Norio KATAYAMA
<katayama@nii.ac.jp>