Visual information retrieval from large distributed online repositories
Communications of the ACM
Digital images and video are becoming an integral part of human communications. The ease of capturing and creating digital images has caused most on-line information sources look more "visual". We use more and more visual content in expressing ideas, reporting, education, and entertainment. With the tremendous amount of visual information becoming on-line, how does one find visual information from distributed repositories efficiently, at least to the same extent as that of existing information
... etrieval systems. With the growing number of on-line users, how does one design a system with performance scalable to a large extent? Visual information is rich in content. The same picture may invoke different responses from different users, at different time, and in different contexts. A picture may have different meanings at different levels, e.g., description, analysis, and interpretation described in  . Visual information may be represented in different forms -still images, video sequences, computer graphics, animations, stereoscopic images, and those for futuristic applications, such as multi-view video and three-dimensional video. Furthermore, visual information demands a large resource of network bandwidth and storage. All these factors have made the indexing and retrieval of visual information a great challenge. Based on our experience in developing visual information retrieval system (VIRS) in the Webbased environment, we present our views of the present and future of VIRS, particularly those providing access to large distributed on-line repositories. We will discuss the role of contentbased visual query, present a taxonomy for classifying existing VIRS, present a working Internet VIRS as a case study, describe a preliminary prototype of Internet meta VIRS, and finally make some bold conjectures about the dominant challenging in developing scalable VIRS for the future Internet environment.