Iris detection method using particle filter and edge directional features

In this paper, we propose an eye detection method that is adaptive to facial rotation, using a particle filter and edge directional features. To estimate the boundary between the iris and sclera or eyelid, the gradient intensities are calculated by 4 directional Prewitt filters in 4 regions. The likelihood used in the particle filter is obtained by averaging the gradient intensities for the specific direction in the 4 regions and the upper eyelid area. Moreover, incorrect detection is avoided by using the facial rotation angle and the eye distance, derived from the positional information of both eyes. From experimental results, the average detection rate of both eyes for facial rotation angles between -15° and 15° is 95.3 %. In addition, the rejection function, reflecting incorrect eye detection, provides a 2.8 % increment in the detection rate for both eyes. The experimental results show that the proposed eye detection method is robust to changes in facial rotation.