JISE

    By maintaining eye contact in visual communication, such as in video conferencing, the feeling of intimacy can be greatly enhanced. Unfortunately, due to the physical constraints on the arrangement of cameras and monitors, it is not easy to preserve eye contact in video conferencing. Most people have tried to solve this problem using hardware approaches, which usually lead to bulky or expensive solutions. In this paper, we propose a real-time software method for preserving eye contact in video conferencing. For each station involved in video conferencing, we use two pre-calibrated cameras to acquire a stereo view of the local participant. Then, we apply disparity-based view morphing to the acquired stereo view to generate the desired view that preserves eye contact with another participant. Our experiments demonstrate that the eye contact problem can indeed be solved in real-time by using our software approach. This is the first time a software approach has been shown to be practical for preserving eye contact in video conferencing.