Our research aims at applying information with regard to a users' gaze to computer interaction. In this paper, we propose an iris detection method that is adaptable to various face and eye movements taken from a facial image. First, our method determines the standard skin color from previous frame. The face region is determined using the color difference from the standard skin color. Secondly, the eye and mouth regions are extracted from the face region by hybrid template matching which consists of edge and color features. Using our method, facial parts can be extracted against individual variation of faces and head movements. Finally the iris regions are detected by using saturation and brightness from the extracted eye regions. Then, the iris positions are detected by the Hough transform. We evaluated our method using facial images containing various positions of faces and eyes. Our method achieved an accuracy rate of approximately 96% for eye region detection, and approximately 92% for iris detection.
展开▼