Classification of facial images was done using the eigen-faces method, which is based on a method known as the principal component analysis.
The average face images of 30 young females aged from 18 to 19 years were prepared, and individual face images were classified into 30 eigen-faces. Then 30 cropped faces of each subject were reconstructed using some of the 30 eigen-faces and the average face. The reconstructed facial images were evaluated as to whether or not they were similar to original facial images by 19 observers.
As a result, estimation scores by observers correlated linearly with the number of eigen-faces. It was considered that the individual face was identified to the cropped face with at least 7 eigen-faces. Furthermore, subjects were classified into 5 groups by cluster analysis with eigen-faces.
