Gaze correction for videoconferencing

Abstract

Communication through electronic link plays an important role in social inter-

action. However, the quality of communication using current video conferencing

systems is severely affected by the lack of eyes contact between each party caused

by the inconsistency between users’ eyes and the camera. Although it has been

solved for the high-end systems, the problem still exists for inexpensive consumer-

level set-ups. This report presents a gaze correction approach for video confer-

encing using a single Kinect sensor. The approach eliminates the gaze problem

but still preserves integrity, and correctness of both the human face as well as the

background scene. The method is applicable for simple home video conferencing:

it uses cheap consumer hardware, and is easy to set-up.