Performance analysis of clustering algorithms for information retrieval in image databases
Tak Kan Lau, Irwin King · 2002
In image databases, a good indexing method makes nearest-neighbor retrieval of images accurate and efficient. Since existing alphanumeric indexing methods are not particularly suitable in image databases, researchers have proposed new methods for indexing by clustering methods. In this paper, we analyze the performance of two unsupervised neural network clustering algorithms, the competitive learning (CL) and rival penalized competitive learning (RPCL), together with k-means and VP-tree for image database indexing. We present some performance experiments to measure their accuracy and efficiency. Based on the experimental results, we concluded that RPCL and CL are good information retrieval in image database.