TheoMedia - Theologie & neue Medien
In this project textual, formal and structural interactions between fundamental theology and media society will be researched. The most important facts which influence the media society of the western world are internet, multi media lifestyle and religious symbolism. This project which is carried out together with the Institute for fundamental theology, KF University Graz and the Institute for Computer Graphic & Vision (University of Technology, Graz) mainly deals with the semi automatically retrieving of religious symbols.
Religious symbols were in former times mainly used only in liturgical events. Today they are media effective prepared and presented. Examples are the presentation in TV of the war ′good against bad′ after 11th September or the Star Wars Trilogy.
In the project a semi automatic digital film footage structuring and analysis according to religious symbols should be implemented. The results of the analysis will be connected with an iconographic database. Joanneum Research will provide its Content Analysis Module, the search infrastructure and already existing annotation tools (as used in GMF4iTV) and will support a postgraduate student of TU Graz.
Goal
The aim of the project is to analyse a film semi automatically. The software which Joanneum Research provides will be improved together with TU Graz in direction of intelligent shot analysis and to extract relevant keyframes automatically. Within these frames specific objects should be defined to analyse their appearence in one or more films. The data thus obtained will be related with an iconographic database.
Result
Extensions to our Semantic video annotation suite (SVAS) which enables film analysts to efficiently annotate video footage. The tool provides several automatic feature extraction methods which support the user with annotation and navigation. The most innovative automatic module included as a plug-in into SVAS is an object recognition and search tool based on Difference-of-Gaussian local key points and computed SIFT descriptors as local low-level features for every frame in the video. This allows the recognition of specific objects in the whole video.
See: Products-Solutions-Services: SVAS
