To simply enable personal authentication by one's face.

An eye position detector 21 detects the center positions of eyes in a face image S0. A normalization section 22 normalizes the distance between the center positions of both eyes and obtains a normalized face image S1. Using a distance between the center positions of both eyes in the normalized face image S1 as a reference, an extraction section 23 sets a trimming frame, having a predetermined norm size to the normalized face image S1, and extracts an input image S2 for authentication. A collation section 14 collates the input image S2 for authentication with authentication images S3 of a multiplicity of registered persons stored in a memory 13 and collates to identify whether the person of the face image S0 is a registered person.