Automatic Calibration of 3D Human Faces

This paper presents a method for correcting the posture of 3D face with unknown position and posture based on the standard face pose. We define a standard posture of face which is not affected by the head gesture. Then the algorithms of principal component analysis and iterative closest point are used to achieve the preliminary adjustment of the target model. For the precise calibration, we use the method of depth values iterative searching to find the point of nose precisely, and propose an algorithm of searching along a straight line to find the ridge line of nose. Ultimately we determine the difference between the posture of the target model and standard posture to obtain accurate calculation of the target model posture.