76740068眼睛接触校正宽基线视频会议系统使用径向基函数.pdfVIP

New Eye Contact Correction Using Radial Basis Function for Wide Baseline Videoconference System Xiaozhou Zhou and Pierre Boulanger Department of Computing Science, University of Alberta Edmonton, Alberta, Canada, T6G1K7 {xzhou3,pierreb}@ualberta.ca Abstract. In this paper, we introduce a novel eye contact correction method for videoconference systems with wide baseline. In this system, assistant cameras are installed on each side of the monitor to help capture the views from left side and right side. A pattern with random dots and Radial Basis Function (RBF) in- terpolation are used to help create precise disparity maps that is then used for re-projection. The interpolated views show a smooth transit among cameras on two sides. The experimental results also demonstrate that the proposed method could be extended to produce more robust and accurate disparity maps than most of the existing algorithms from regular stereo images. Keywords: eye contact correction, immersive videoconference, computer vi- sion, stereo, radial basis function interpolation 1 Introduction Videoconference system has been a hot research topic in recent years as it creates a true sense of immersion/presence between each participant by keeping eye contact and displaying participants at their real scale. It has been widely used in business, education, and personal entertainment and has been commercialized by numerous companies such as SONY, CISCO systems, and DVE. The key element of immersive videoconference systems is to maintain the correct eye contact as it carries non-verbal clues that are essential for emotional communications. In