Purpose

Hypervideo is video-centric hypermedia that offers a deep integration of audiovisual material in heterogeneous multimedia documents allowing many interesting audiovisual experiences.

The purpose of the CHM project is to introduce the hypervideo as a new domain, providing a data model that formally describes such documents and their behaviors besides the conventional hypermedia/multimedia ones.

A high level representation of hypervideo documents is thus defined, through :

• A component-based modeling and
• An annotation-driven content and structure description.

>> This allows hypervideo to be considered as a video-centric document that inherits from both multimedia and hypermedia fields.

What are hypervideos?

Wikipedia defines hypervideo as a displayed video stream that contains embedded, user-clickable anchors, allowing navigation between video and other hypermedia elements.

Hypervideo results from the combination of interactive video with traditional hypertext and thus, it is associated with several definitions depending on the emphasized point of view. While some works emphasize the hypermedia aspect of adding information into digital video so that users can activate video hyperlinks and get access to additional information in form of texts, images or even other videos, other authors point up the storytelling aspects through hyperlink activation to dynamically create a non-linear and user-defined navigation path in the document. Hence, some authors describe hypervideos as being just the inclusion of clickable video on hypermedia, others consider it a novel multimedia document form to investigate.

Issues with Hypervideos

• Many authoring and reading concerns are particularly associated to hypervideos:
  • The continuous nature of video results in an information distribution over space and time, making data seizing, access and processing a complex task.
  • Common hypermedia concerns might be more pronounced in hypervideo, with important risk of overstraining the cognitive capacities of the user and putting him under time pressure during navigation.
  • Important rhetorical and aesthetic challenges are involved with regard to different user attitudes towards the nature of such documents.
  • The potential of increased cognitive load might lead to user disorientation, which may mean "lost in space" but also "lost in time", and proper information delivery inefficiency.
  • Hypervideo links may have durations - for a very time limited link opportunity, the user has a restricted lapse of time to activate it -, several links can coexist in space and time - link awareness alternatives have to be faced to promote link anchors existence perception - and the presented content changes in time.
• The existing systems representations are mainly technically driven and do not endorse a well defined hypervideo data model. Many of these systems are specific for particular needs and thus lack of generality. Features like semantic description and content enrichment, video fragmentation and composition, robust addressing mechanisms, linkage and fragment accessing and so forth, are not widely granted.

Hypervideo and Annotations

Video not being structured data-wise, the demand for added semantic description and content enrichment stimulated the emergence of video annotation efforts towards the disclosure, explanation and augmentation of the carried knowledge. The adoption of hypermedia terminology to video annotation is often referred to as hypervideo, since it resolves the linearity of video structures and so creates a non-linear information space. Most of hypervideo editors use the concept of annotation and provide means to add such data to video. Systems like HyperCafe, HyperSoap, VisualShock movie and HyperFilm are considered to be the first research projects that integrate video annotation as a core concept while Hyper-Hitchcock allowed viewers not only to interact with annotated videos but also to create their own annotations and share them with others over the Internet in real-time. MPEG-7 offers a comprehensive set of audiovisual description tools to create such metadata. Adivi, Advene, Anvil and IBM's VideoAnnEx and MediaDiver provide the functionality to annotate videos by text, links or any multimedia content and to generate rich video-centric presentations.

Annotation-driven approaches for hypervideo design define various access granularities by allowing video fragment definition with no length restriction. Fragments can be arbitrarily overlapped or nested, associating various data and multiple views to the same audiovisual information unit. Thanks to the data separation from its possible visualizations, maintaining the document structure is eased, independently from the audiovisual stream. This improves security management and enhances collaboration options by requiring to update or exchange only the annotation structure. Moreover, with semantic content annotations, data analysis and querying can be performed at a level that current image processing and computer vision techniques cannot achieve.

---