A fast eye location algorithm based on geometric complexity

In a frontal face-image, the major facial components (eyes, eyebrow, nose, mouth) have some particular geometrical characteristics. By properly introducing some geometrical complexity concepts, this paper presents a method to quantitatively analyze the geometrical characteristics around eye area. In this way, a disoperative face-image can be effectively detected in advance. Then, in a normal frontal face-image, double eyeballs can be located precisely. In our experiment, total 520 simple-background frontal face images were tested. 100% disoperative faces (32 images) were successfully detected, and 99.4% the accurate rate of precise eye location was achieved. Compared with other methods, this algorithm can locate double eyes with less computational expense and more precision, suitable for embedded real-time applications.