Distributed image classification based on high-order features

This paper presents a new high-order feature for image classification based on distributed hadoop implementation. The proposed method firstly extract SIFT features from image, and then divide image into multiply grids. In each grid, the strongest SIFT feature is regarded as major feature while the other SIFT features as minor features. The high-order feature is composed by major feature, minor features and angels between features. Finally, a distributed hadoop implementation of image classification based on high-order feature is proposed. Through experiments, our proposed approach performs favorably while compared with two well-known algorithms in a benchmark dataset.