A feature region is extracted from the image region of an object, in accordance with the luminance distribution of pixels adjacent to a predetermined pixel that exists in the image region of the object (S31 to S34). A gravity-center correlation value is calculated, which shows the correlation between the gravity center of the feature region and the gravity center of the image region of the object (S35 and S36). The orientation of the object is determined from the gravity-center correlation value (S37).