Structural element candidates, estimated to be predetermined structural elements of a predetermined subject, are detected from an image that includes the subject. The subject that includes the structural element candidates is detected from the image in the vicinity of the detected structural element candidates. The characteristics of the structural elements are discriminated from the image in the vicinity of the structural element candidates, at a higher accuracy than when the structural elements were detected. In the case that the characteristics of the structural elements are discriminated, the structural element candidates are confirmed as being the predetermined structural elements.