Democratic integration for segmentation in a 3D videoconference system

In this paper we present an algorithm for video segmentation and background suppression in 3D videoconferencing. It is based on a new data fusion method called Democratic Integration. This technique has been developed to make decisions on basis of multiple sensor signals and information channels with temporally changing reliability. In this context, our contribution particularly shows how this method can be utilised to improve object segmentation in a real-time 3D videoconference application. Results will be shown on the basis of current implementations of an immersive 3D videoconferencing systems which is under investigation at Fraunhofer-Institut für Nachrichtentechnik, Heinrich-Hertz-Institut (HHI).