A method and a system for simply and automatically detecting the positions of eyes, mouth, and nose in a face image with high reliability are provided. A plurality of grayscale images with gradually varying lightness are formed from a face image. Then, a process (fade-in process) is performed which detects pixel block areas that gradually appear in a face area in the grayscale images as the face area that is in a fade-out state at high lightness fades in and is brought into a fade-in state at low lightness. The detected pixel block areas include paired pixel block areas appearing as a pair. The positions of the eyes are determined based on the number of appearance of the paired pixel block areas over all the grayscale images. The positions of the mouth and nose are determined based on the positional relationship between the eyes and pixel block areas which are detected by the fade-in method and in which gradually appearing pixels no longer grow.