A Novel Fast Massive Data Retrieval Method

In this paper, we proposed a novel fast massive data retrieval approach, and utilize this approach in an image retrieval system. Framework of the proposed massive image retrieval system is proposed in advance, and key part of this system is the searching algorithm. Considering the efficiency of traditional K-means is not satisfied by us, we present an improved clustering algorithm to solve this problem. Particular, we modify the K-means algorithm with the discrete function of a specific level's histogram value to obtain the cluster centroids of all images. To testify the effectiveness of this method, we collect four image datasets to make performance evaluation. Compared with the traditional K-means algorithm, the proposed method can obviously promote the accuracy of data retrieval and cut down the time cost as well.