The research of descriptor extraction accelerated method based on image content retrieval

Low-level image feature extraction and index is a necessary step for content based image retrieval, but traditional feature extraction method for large size and large capacity images is usually time-consuming. However, this kind of image retrieval is the trend of the development of Internet image retrieval. This paper proposes a random feature extraction method based on compressive sampling which selects 1% of the image pixels through the random mask to extract low-level feature vector. In this paper, the method is used in MPEG - 7 dominant color descriptor (DCD) and edge histogram descriptor (EHD) extraction, and discusses the influence of different masks on image low-level feature extraction accuracy. The experiment proves that this method can effectively improve the efficiency of the low-level feature extraction without affecting the extraction accuracy.