An image forming apparatus includes a document reading unit configured to collectively read a plurality of source documents placed on a document table, a reference document image selector that selects, as a reference document image, an individual image representing the source document located closest to a position corresponding to a far left corner of the document table, a first relative position detector that detects a relative position between the reference document image and each of the individual images other than the reference document image, and a first pair identifier that identifies, as an image pair representing front and back faces of the same source document, a pair of individual images, the respective relative positions of which, from the reference document image in one of the both image data and from the reference document image in the other one of the both image data, are closest to each other.