This paper presents a fast method for detecting the eyes and the mouth corners on a facial image with significantly varied head poses. The described scheme only reasonably assumes that the facial image should be thresheld well enough that the hair and eyes can be distinctly separated from the cheeks. It first detects the two irises by using the Hough Transform together with the constraints on the facial symmetry and similarity then roughly binds themouth region according to the anthropometric standards

and finally utilizes the integral projections to locate the two mouth corners. This method has been tested on more than 250 face pictures with different illuminating conditions and varied head poses

and showed a quite satisfactory performance both on speed and accuracy.