A hybrid classifier and template matching method for eye detecting in real-time video

Eye detecting is an important part of Computer Vision. Even if there are a large number of the methods for eye-detecting nowadays, most of them are vulnerable to different light conditions and complex environments. This paper proposed a novel Classifier and Template Matching method for eye detecting in real-time video. The presented method includes three parts. Firstly, read the video from the camera frame by frame, we utilize the existing strong and efficient classifier which is produced by Harr-like features to find out the human face in each frame. Secondly, shrink the position of eyes based on our experience in the detected face we have got. Thirdly, get the exact position of eyes with template matching algorithm. Experiment results demonstrate that the method is robust for detecting human eyes in most situations.