Performance of comparative reading of two or more images that are to become the objects of comparative reading is improved with respect to conventional means. An energy subtraction image forming means (structure-emphasized image obtaining means) obtains an energy subtraction image based on each pair of energy subtraction-use original images, and a structural-position relation obtaining means obtains the correlation between structural positions within each pair of energy subtraction image obtained thereby. Based on the correlation between structural positions of the energy subtraction images, an image position matching means matches the positions of the structures between each pair of energy subtraction-use original images.