A plurality of feature points are extracted from an inputted image, and a feature amount is computed with regard to the feature point thus extracted. A prescribed plurality of alterations, i.e., a rotation, an enlargement, or a reduction, are applied to the inputted image, a plurality of altered feature amounts are estimated with regard to the feature point, and an image is retrieved from among stored images that are stored upon a storage apparatus and that is most similar to the inputted image, in accordance with the plurality of the estimated altered feature amounts.