IMAGINATION - Image-based Navigation in Multimedia Archives
IMAGINATION aims to ease knowledge gain from digital cultural and scientific resources by enabling navigation through images and their context.
- Project type: STREP (Specific Targeted Research Project)
- Start date: 1 May 2006
- Duration: 36 months
- EU funding: € 2 130 000
- Number of partners: 8
- Project coordinator: Forschungszentrum Informatik an der Universität Karlsruhe
- Contact: Mr. Clemens van Dinther
Point of departure for IMAGINATION is that the human brain is more efficient in processing information in visual than in textual form and that knowledge is generated not from single information items but from information in context. The project's research goal is thus to explore how information systems could become more performing in supporting image-based and context-aware interaction.
Currently available methods for searching and accessing large repositories based on image content still present shortcomings in usability and performance. This is due the limited feasibility to generate semantic descriptions for images. Based on recent technologies such as RDF and OWL, it is possible today to define semantic metadata for images, and use it for the three main tasks of information systems: to find relevant resources, to support user navigation and to display contextual information. However, existing image segmentation and image recognition algorithms cannot yet reach a level of precision adequate for automatic annotation.
To solve this problem, IMAGINATION proposes to make use of the fact that resources are not stored alone in a repository, but they are usually embedded in and together with other resources. IMAGINATION plans to develop a browsing environment that will help logically and predictably organise a database of images and texts and facilitate the user's search tasks through efficient and automatic browsing tools.
Work towards this goal will consist in combining and further developing recent techniques in the fields of text-mining, image segmentation and image recognition. This will result in a set of new technologies and tools that support the manual, semi-automatic and automatic annotation of images with high-level semantic metadata. This high-quality metadata will be used to visualise the context of resources (images and texts) stored in knowledge spaces, and thus make them more meaningful for users.
IMAGINATION's work will create the basis for a new type of multimedia archive: applying the paradigm of the web, where navigation is based on interlinked text, the project will create an image-web, where images and parts of images are semi-automatically interlinked and browsable.