Abstract
A fast algorithm is described for detecting the human face symmetry rate, an important feature of the human face. Symmetry rates can be obtained simultaneously not only for the full face, but also for each feature (eyes, nose, mouth, etc.). This algorithm rejects the nose's shadow, which is an obstacle to obtaining the symmetry rate. First, the target face is divided into six vertical blocks, and the center of gravity of each block is calculated. The principal axis of the face is determined from the obtained gravity centers by using linear approximation (minimum mean square error estimation). Rejection of shadows and rotation of the principal axis to make it perpendicular to the horizontal axis are done next. Finally, the symmetry rate is calculated. Rejection of any moles on the face due to the shadow rejection is prevented by using a circular rate. This algorithm includes real part-only synthesis in order to obtain a high-efficiency fast calculation. Simulation examples showed that it is effective for quickly determining the symmetry rate of a human face.
