Stereo camera based real-time gaze correction method by matching and blending eye regions

Abstract

In video-conference, the camera is placed at the top of the screen and the user looks at the other user's image on the screen. This misalignment of eye gaze direction degrades the user's experience. To solve this problem, this paper presents an eye gaze correction method using vertical stereo camera. The proposed method matches the eye regions in two images captured by stereo camera and generates new eye regions by alpha-blending these two eye regions. To obtain a gaze corrected image, we replace eye regions in top image with new eye regions. The proposed method requires fewer computations than conventional methods using stereo camera. Experiment results showed that the proposed method could generate eye gaze corrected images at 14 frames per second using images captured simultaneously by top and bottom cameras with the resolution of 640×480.