Improved Eye Gaze Detection using Contour Technique

The main objective of this paper is to speed up the gaze point detection with reliable accuracy. Even though the accuracy level of gaze point detection is good in the existing tracking systems, it also includes some known limitations such as restricted head poses, poor eye blink ratio and the limited speed of eye detection. In this paper, we have also tried to improve and overcome the above stated problems by making use of the Local Contour Sequence technique during the iris detection process, which is invariant to translation and rotation of major head poses, various facial features and maintains the very minimal gaze error on continuous eye blink. Here, the initially detected face region is considered as a collection of pixels and from those contour pixels we need to distinguish the smooth and exact eye object from those lighter and darker background pixels. For this main use case, we made use of the local contour sequence in our program - which fastens the eye detection.