Accelerated kmeans Clustering using Binary Random Projection

Abstract

Codebook has been widely used for image retrieval and image indexing, which are the core part of the mobile visual search. Building vocabulary tree is carried out offine, because the clustering of large number of training data takes long time. Recently proposed adaptive vocabulray trees do not need the offline traning, but it suffers from the online computation. The necessity for clustering high dimensional large data has arisen in offline or online training. In this paper, we present a novel clustering method to reduce the burden of computation without losing the accuracy. Feature selection is used to reduce the computational complexity in high dimension of data, and ensemble learning model is used to improve the efficiency in large number of data. We demonstrate the proposed method outperforms the-state-of-the-arts in terms of computational complexity on various synthetic and real datasets.