Study of content-based image retrieval using parallel computing technique

Abstract

In this paper, we implement a parallel technique to a content-based image retrieval system and investigate the influence of different hardware devices on the retrieval system performance. Our system is developed on cluster architecture. The feature extraction and similarity comparison of visual features, which are widely used for the content-based image retrieval, are realized by using the parallel computing technique. Therein, the feature vector, which identifies one image uniquely, is composed of several different color features. Numerical experiments have demonstrated that the parallel computing techniques can be applied to significantly improve the performance of retrieval system.