A Solution to the Eye Contact Correction in Tele-presence Systems

• Author / Creator

  Zhou,Xiaozhou

• With the development of globalization, numerous tele-conferencing systems have been developed to shorten the distance between people. Tele-presence is one of these systems which can broadcast high-quality audio and video to remote sites trying to give the illusion to the participants that they are co-located in a common virtual meeting room. In order to provide this illusion, one main criterion is to maintain eye contact. This is important because eye contact can convey more information than languages sometimes and correct eye contact is one of the essential factors which make the participants feel like they are talking face-to-face. However, some of the successful methods to eye contact correction problem in tele-presence systems need special equipment which is either large or expensive. In this thesis, we propose a software-based solution to the eye contact correction problem. This solution uses depth-based view interpolation to create virtual views from a network of cameras located around the display screen. Using this method, users have the freedom to adjust their viewpoints to allow eye contact or to focus their attention on important regions.

  To improve the illusion of being in a common meeting room, we propose a new image matting algorithm to extract a participant from the background that can deal with hair and small details. This is essential as any small imperfections in the matte will destroy the illusion of being in a same virtual meeting room.

  In the thesis, we also explore how to overcome the uneven illumination found in tele-presence rooms. We present solutions to the illumination invariant stereo matching problem to create better disparity maps. Taking the advantage of the parallel structure of local stereo matching, the GPU implementation of the proposed solution can run in real time. Intensive experiments demonstrate the effectiveness and efficiency.

  In general, our main goal in this thesis is to develop core technologies that will eventually be used in future tele-presence systems.

• Subjects / Keywords

  • Stereo matching
  • Eye contact correction
  • Image matting
  • Computer vision

• Graduation date

  Fall 2013

• Type of Item

  Thesis

• Degree

  Doctor of Philosophy

• DOI

  https://doi.org/10.7939/R34F1MV8G

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.