Abstract
Our society is creating, storing and using an ever increasing volumes of digital multimedia content. We have access to hundreds of billions of images and videos on the Internet, in the archives of professional content creators and owners, and in the personal libraries of home users. In this context, content-based identification of images' and videos and visual search capabilities are essential enablers for a range of applications that require fast, robust and efficient algorithms. The interoperability of the technologies and systems is also an important consideration. During the lecture Miroslaw Bober will review the latest advances in content fingerprinting and visual search and how they impact related standardisation efforts within the MPEG group. In particular will present the MPEG Visual Signatures, which include the Image Signature and Video Signature content description tools. The Visual Signatures are designed specifically to enable fast and robust identification of near duplicate or derived (modified) visual content in large-scale databases. We will also look at the latest addition to the MPEG family of standards: Compact Descriptors for Visual Search (CDVS). Achieving a state-of-the-art identification and recognition performance, the applications are numerous and include rights management and monetization, distribution management, usage monitoring, and professional or personal database management.