InfoNames: An Information-Based Naming Scheme for Multimedia Content

Abstract

Recent proposals have argued for data-centric mechanisms that decouple data delivery from the sources of the data and the transfer protocols. We take this idea to its logical completion and argue for enabling content distribution schemes to name and query directly for the underlying 'information'. The motivation for this information-aware design is that we see a proliferation of diverse producers of multimedia content offering varying presentation modes and significant heterogeneity in the operating conditions of Internet-enabled devices that seek access to such multimedia content.

In addition to decoupling content from available sources and transfer protocols, information-aware names, or InfoNames, explicitly decouple the information from content presentation. Thus, it enhances availability and allows users to maximally leverage sources of multimedia content offering diverse presentation formats. Further, users and providers benefit from having additional flexibility to dynamically adapt content delivery depending on application and network constraints.

In this report, we address the first challenge to realizing this idea - How should we 'name information' in multimedia content in a way that is invariant across presentation formats and can be consistently generated/verified without relying on a centralized authority? We leverage techniques from the multimedia and computer vision communities, and propose a set of algorithms for naming, and comparing InfoNames. An extensive evaluation of the proposed schemes on a controlled dataset of images and videos is presented. In addition, an 'in-the-wild' study with a set of videos download from Youtube is also described.