Virtual Reality is increasingly entering the mainstream and finding a place in consumer households, from high-end devices such as the Oculus Rift and HTC Vive down to more affordable mobile solutions like Samsung Gear VR and Google Cardboard. As VR becomes prevalent, traditional applications are moving to VR to take advantage of the potential for increased interactivity and immersion. In consumer surveys, a large percentage of Millennials responds positively towards VR. However, VR is still cost prohibitive for many users, as well as impractical in many situations, for reasons of hardware requirements or social acceptance. Not much work has gone into how to represent a traditional webcam user to VR users in a way unique to the medium. This project explores novel methods of displaying a 2D webcam feed to a VR power user with minimal setup and overhead for the webcam user.
The first approach explored is segmenting the user out of their webcam video feed in real-time and placing them onto a planar polygon in VR. This process consists of efficient real-time segmentation of the webcam feed and the completely distinct problem of how to display that segmented image in VR. The second approach explored is attempting to turn the 2D image into a realistic 3D avatar representation of the user. This too requires tackling an image processing portion as well as HCI decisions about representation. Finally, after exploring the different solutions, we address their potential limitations as well as practical aspects such as the hardware/software architecture for a system that could handle such video calls.