Content-Based Semantic Retrieval of Video Documents, Application to Navigation, Research and Automatic Content Abstraction.

 

 CNCSIS project RP-2

Introduction: What is the project thematic ?

At the present time, the access to mutimedia information is crucial and takes part of our day to day life.

Thanks to the technological evolution, the amount of multimedia data to interact with, i.e. text, sound, image, video, etc., has become increasly large. Among these categories a particular scientific and economic interest is shown by the video documents.

The actual issue is not the lack of information, but instead, the dificulty of accessing large amount of information. The existing solution are the indexing systems. Indexing data means to annotate its content to facilitate the retreival task.

An example of automatic understanding of the video content using image processing.

 

Among the existing indexing systems the most interestings are the ones that use the semantic content annotation.

Therefore, the video documents are retrieved in a human-like manner such as using a common language: find me the action movies or find me the goal sequences. A semantic indexing system requires the complete understanding of the video content, which  unfortunately, it is not totaly possible with the actual scientific progress.

The innovating part of our project consists precisely in pushing forward this research field which is still at his early years. We are proposing to develop new methods for the automatic understanding of the video content, all in order to facilitate the retrieval task. These methods should exploit together different sources of information, such as colour, texture, motion, temporal structure, text, etc.

We are also proposing to develop new methods for smart automatic content abstraction and content representation (semantic graphical user-interfaces) which are mandatory for the research task.

The final target is the development of an indexing system which should provide video content retrieval through a natural and intuitive language making it suitable for all categories of users (professional or common people).