ABSTRACT
It is very important to extract the eye features accurately in the facial analysis and processing. Nowadays, most methods to extract these features are used in static images. These methods can extract the eye features effectively when the eye opens widely. However, in some application fields which need real-time environment, for example, fatigue detection of driver, they need to extract the eye feature not only real-time but also accurately when the eye opens narrowly [1]. To solve these questions mentioned above, this paper presents a method based on gray information and Harris corner detection. Firstly, the method locates the facial area accurately based on skin-color model algorithm [2]. Secondly, we do the size normalization and grayscale equalization for the face sub-image and find out the detection windows of left and right eyes. Thirdly, we use difference projection to locate the eye area roughly in the eye detection windows. Then we locate the pupil position accurately in the eye areas obtained from the previous step by gray statistical information and detect the position of eye corner by Harris corner detection algorithm. Finally, we portray the contour of eye with the method of curve fitting according to the position of the pupil and eye. This method extracts the outline of eye accurately and needs few computations, and it can be used in the real-time environment.
