Eyes localization in an infrared image

This paper proposes a novel quick method to locate a person´s eyes in an infrared image. First, we use the homomorphic filter to enhance the contrast of image, segment the human face through iterative threshold choosing algorithm and integral projection function. With the eyes and face geometry knowledge, we get the eyes region roughly. Second, we use the RAMF(Ranked-order Adaptive Median Filtering) method and iterative threshold choosing algorithm to get the human eyes region accurately. Finally, we get the eyes position according to the centroid of the eyes region. Experiment results show that the proposed method is simple, fast and reliable.