PURPOSE: To lighten the burden on an operator by automatically generating format information.

CONSTITUTION: A document 1 is read by an image acquiring means 20 and its document image is stored in an image storage means 30 in the form of image data. The positions of segments and/or the positions of characters included in the read document image are analyzed by an image analyzing and character recognizing means 41. A collation and retrieval means 42 collates the image analysis and character recognition result of the image analyzing and character recognizing means 41 with a format generation rule, retrieves image analysis and character recognition results matching the format generation rule by read fields and determines the positions of those read fields, and generates the format information.