iGLANCE - interactive Genius Look At Numerous Contemporary Events

The development of a new generation of displays that allow perceiving depth without glasses requires new algorithms of image and video processing. 3D video and free-viewpoint video provide new sensations for the users and will create new markets.

Generating free-viewpoint 3D video is a challenging topic in computer vision. This technology enables the user to view a scene from a freely chosen position: for example, the user can watch a concert or football game from the viewpoint he prefers.

We concentrate on the algorithms for generating intermediate 3D video views, given several 3D views of the scene. These algorithms include multi-view video interpolation and depth estimation.

3D video processing is increasingly important for medical purposes as well as in the consumer electronics domain. In the medical domain, 3D data can be generated for several imaging modalities, and visualizing this data on a 3D screen with a free-viewpoint navigation will help to optimally navigate instruments during operations, making the operations minimally invasive.

Working in close collaboration with internationally renowned industrial partners and research institutions, while developing software and hardware simultaneously, we hope to achieve not only scientific but also R&D relevant output.

ViCoMo - Visual Context Modelling

Video content analysis is increasingly important for security, medical, logistics and other domains.

ViCoMo is an international ITEA project jointly led by Eindhoven University of Technology, Philips, INRIA, Thales, Cyclomedia Technology, and other academic and industrial partners. The project focuses on the following aspects of multi-view video content analysis:

human behavior;

context of events (for example, analyzing a training session of a group of medical professionals in a context of their interactive behavior can help to determine whether the people act individually or as a group and whether there is a cooperation or panic);

behavior of objects that are critical (examples include observation for surveillance, cargo operation control in port terminals, etc.);

modeling interactions between events or people or objects;

fast solutions and real-time systems.

The research challenge comes from the use of multiple views of the same scene (multiple cameras) and detecting, recognizing and retrieving objects and events from the recorded multi-view video sequences. This involves advanced video content analysis, 3D reconstruction, object tracking.