Whether a portrait photography mode has been set is determined. A first threshold value is set when the portrait photography mode has been set and a second threshold value that is smaller than the first threshold value is set when the portrait photography mode has not been set. A cross-correlation coefficient between the image of the subject and a template image is calculated. The calculated coefficient and the set threshold value are compared and the position of the image of a face is detected in the image of the subject. When the portrait photography mode has been set, it can be construed that the subject contains a person and therefore the position of the face image can be detected comparatively accurately by increasing the threshold value.