ABSTRACT
This paper presents a Composite Cross Correlation technique for detecting eye in varied head pose of face images. The proposed composite template consists of two single templates; the first template is created by taking the left half of a left eye and the second template by flipping the first template horizontally. We suggest other composite templates of entire face template to determine how many eyes should appear, one in profile face image and two in frontal face image. The experimental result was conducted on Psychological Image Collection at Stirling (PICS) database. The database has face images with five head poses in the range -90°, -45°, 0°, 45° and 90°. The results obtained using the proposed technique shows significant improvement in terms of detection rate compared to single cross correlation or using eye template of entire eye images.
