An image processing apparatus has a read image acquisition section and a specific image region extraction section. The read image acquisition section acquires read image data which is obtained by reading a front face and a rear face of a document. The specific image region extraction section extracts, as a specific image region, from among local regions existing at corresponding positions of the front face and the rear face of the read image data, a local region in which an amount of difference of at least one of hue degree, brightness degree or saturation degree between the front face and the rear face is within a predetermined range.