SEMINAR: Implementing a robust teleconferencing system using multiple depth-sensing cameras

Afsaneh Rafighi
M.Sc. Thesis Proposal
Supervisor: Dr. Oscar Meruvia-Pastor

Implementing a robust teleconferencing system using multiple depth-sensing cameras

Department of Computer Science
Thursday, October 16, 2014, 1:15 p.m., Room EN 2022


Abstract

The introduction of depth sensing cameras has resulted in huge breakthroughs in computer vision and robotics. RGB-D cameras provide RGB images along with their depth information. The added depth information has resulted in many innovations. They are used in robotics, 3D map reconstruction, human computer interaction, 3D modelling, visualization, telepresence, gaming filming industry. One area of research which has become attractive lately is using depth cameras in video conferencing systems to convey the feeling of 3D to users. These cameras, however, have their own limitation in field of view and suffer from depth inconsistency.

The idea of having multiple depth-sensing cameras in a video conferencing system to act as one camera is the topic of interest for this research. Having multiple cameras helps to enlarge the field of view and observe the views which are occluded from one camera. It also helps to find the missing depth values from one camera with the help of the other, using the overlapping regions. Moreover, it could be used to build a virtual space giving the sense of the real presence of the person at the other side.

There are video conferencing and teleconferencing systems being presented so far that use single or multiple cameras at each point. However, all of them require a manual calibration of depth sensing cameras and will fail if cameras are moved. Calibration process is time consuming, and in case one camera moves, they need to be calibrated again.

In this research we are aiming to build a robust teleconferencing system using different layers to acquire and compress data, transmit it through network and once received at the receiver’s end, decompress it and apply registration to merge the point clouds. This will be a robust system that users are sending and receiving data simultaneously without the need of pre-calibrating the cameras. To do so, we intend to use the previous work of Sahand Seifi (M.Sc. cadidate) et al. who presented a registration method called DeReEs.

 

Contact

Department of Computer Science

230 Elizabeth Ave

St. John's, NL A1B 3X9 CANADA

Tel: (709) 864-2530

Fax: (709) 864-2552

becomestudent@mun.ca