A gaze tracking scheme with low resolution image

To estimate the eye gaze accurately, a high-resolution and good quality image of the eye region is essential. Therefore, one or multiple narrow angle cameras with long lens are commonly used. In case of poor image resolution or absence of narrow angle focusing, the gaze tracking would be inaccurate. This paper looks into the possibility of using low resolution wide angle camera to track gaze. In order to combat the limitation of poor eye-region image which might only be defined by a few pixels, a precise corneal reflection vector estimation based on local image reconstruction method is proposed. By using 2D bilinear interpolation, the local eye-region image is enhanced for precise reflection vector estimation. Although the accuracy of the proposed method can barely be compared with that of the state of art techniques, it is yet noteworthy that the performance gain has been achieved with low resolution images only, indicating the possibility of tracking gaze with low-cost simple cameras.