The accuracy of eye tracking using image processing

  • Authors:
  • Gintautas Daunys;Nerijus Ramanauskas

  • Affiliations:
  • Siauliai University, Lithuania;Siauliai University, Lithuania

  • Venue:
  • Proceedings of the third Nordic conference on Human-computer interaction
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

Eyegaze is one of an alternative input methods. Nowadays, the videoculography (VOG) is the most suitable for this purpose. Eye pupil center position is necessary for evaluation of gaze direction. The novel pupil center co-ordinates extraction method was proposed. The method exploits averaging of pupil edge co-ordinates in the same scan line. Obtained points are aproximated by line. Pupil center is estimated as intersection of vertical and horizontal lines. Accuracy of method was tested using synthetic eye images. The novel method accuracy is similar to one of approximation by circle method, when an eye is looking straight a head. The proposed method has better computational effectiveness and smaller errors in tertiary eye positions. Errors for both methods increase versus noise standard deviation with the same slope. However the influence of noise is small, if region between iris and pupil is sharp.