A document image processing device can correct the location of a document of an arbitrary format. The reference position of a document image is designated previously by a reference position designation unit. The document image is input per page and stored. A predetermined pixel block extraction unit extracts the predetermined pixel block appearing commonly on each page, and the coordinate value of the predetermined pixel block is supplied to a difference extraction unit. The difference extraction unit calculates the difference value between the position of the extracted predetermined pixel block and the reference position designated by the reference position designation unit, and supplies the difference value to an image shifting unit. The image shifting unit shifts the document image of that page based on the difference value, and corrects the location of the document image. The predetermined pixel block used for position correction can be set arbitrarily.