Analyzing large image databases with the evolving tree

Analyzing large image databases is an interesting problem that has many applications. The entire problem is very broad and contains difficult subproblems dealing with image analysis, feature selection, database management, and so on. In this paper we deal with efficient clustering and indexing of large feature vector sets. Our main tool is the Evolving Tree, an unsupervised, hierarchical, tree-shaped neural network. It has been designed to facilitate efficient analysis and searches of large data sets. Comparison to other similar methods show a favorable performance for the Evolving Tree.