Focus of Attention in Video Conferencing

Abstract : In this report we present an approach to low bitrate video teleconferencing by focusing attention on important information. We show that by selectively degrading the quality of less important regions, more important regions can be sent without loss of quality but with greatly reduced bandwidth requirements. Low bitrate transmission for real-time video delivery over a dynamic network is achieved by region blurring and cropping. A prototype system has been developed to demonstrate the concept. We assume that a human facial area is the most interesting area in the system. The system can automatically focus its attention on a given face and its adjustable surrounding area. The selected area is then fed to the coding system and sent to the receiver. The selection function of the system is fulfilled by a real-time face tracker. The face being tracked can be selected by a mouse or finger pointing if a touch screen is used. It can track a person's face while the person moves freely (e.g., walking, jumping, sitting and rising) in a room. Based on the information provided by the face tracker and the network traffic, a window surrounding the face can be determined. The window size can reflect the network traffic. The image outside of the window will be either cropped or blurred. The preprocessed image is then fed to a tele-conferencing software package-vic, a real-time multimedia application for video conferencing over the Internet. The experimental results show significant savings of required bandwidth for video subjected to the changes.