Multi-camera Computational Video Architectures
Computational video and vision systems allow for new applications going far beyond digital TV, IP-TV and traditional camera based systems. EDM/UHasselt is renowned for its groundbreaking research in the area of software systems for computationele video en multi-camera 3D systems. The execution of video algorithms is extremely computation intensive. This is being determined by the image resolution (quadratic dependence on image dimensions), the amount of frames per second, the color accuracy, the amount of cameras deployed etc.. The image processing algorithms used, often based on complex arithmetic, of course have an even higher impact. Even when the most powerful PC┌s are being used, one is still limited in the possibilities and the video quality that is being targeted. This is especially true if real-time processing is required. The underlying research proposal aims at the development of new application oriented computational video hardware/software architectures. Hereby new video- and vision algorithms and systems can be implemented that could previously not be realized. There are several applications for the result of this research: 3D television, virtual camera systems, high quality/high speed robot vision and visual inspection systems, visual traffic monitoring and management systems. In the automotive sector, intelligent cameras will enable lots of new applications: lane and traffic sign detection, tunnel detection, passenger monitoring, road hazard detection, etc. In all of these applications real-time application oriented computational hardware/software vision systems enable large advantages as well as new unexplored application areas.
Period of project
01 January 2009 - 31 January 2012