Localization and extraction on the eyes, nose and mouth of human face

The traditional gray projection function was mainly utilized in eye localization. In this paper, an optimized function is proposed to simultaneously localize eyes, nose and mouth. This improvement is implemented through three steps. Firstly, by calculating the derivation of the gray projection curve, the stationary points region of eyes (nose or mouth) can be obtained. Secondly, determine the boundaries of those stationary points regions by setting appropriate thresholds. Thirdly, extract facial features, such as eyes, nose, and mouth from the original images in accordance with those boundaries of the respective stationary points region.