Image Retrieval In The Print Media

With the development of the mobile Internet, the print media technology is undergoing tremendous changes. Combined network technology with image processing technology promotes the digitization of print media. In the print media, the image is an important way to transmit information and an important factor for attracting users and readers to interact with the media. Users take some special pictures through the mobile terminal, such as car standard, luxury brand logo, web logo and so on, and then carry out the image information, track the dynamic and analyze the communication using image processing technology. In this process, the key technology is image retrieval technology. Therefore, how to accurately retrieve the target image from a large number of images becomes a meaningful research topic.The paper analyzes the development and existing problems of image retrieval technology in current days firstly. Secondly the paper introduces the key technologies of image retrieval based on local feature, including image preprocessing, feature extraction and feature matching. Also we describe several common methods of local feature extraction, especially SIFT algorithm. Then, on the base of all the technologies, we combine the scene with the problems of the algorithms and improve the key technologies, SIFT-PCA, which solves the problem of time caused by high dimension. Finally, the system of image retrieval on the print media has been implemented based on the key technology improved. Also, we compare the technology improved with others in terms of performance and efficiency through experiment. The results show that the improved SIFT-PCA algorithm has both good performance and high computing speed and it is ideal for image retrieval in print media.