A method of providing a digital image with a unique, machine readable, code image is presented. It is called "Full Spectrum," because the method uses Fourier transform techniques to embed the code in a wide range of spatial frequencies. The changes made to the original image by the encoding process are meaningless to the observer, and, by proper choice of the embedding parameters and the resolution (reproduction size) of the marked image, they can be made imperceptible. Full Spectrum has some intrinsic advantages for application to printed security structures in a document authentication and identification environment. Using its mathematical properties, the method is shown to be invariant to shifting and robust to cropping, which enables the code image to be reconstructed from a recording of the document with arbitrary position and size. Techniques to deal with possible rotation and scaling of the recorded image with respect to the (printed) original are elaborated. Finally, a general method is developed to match the reconstructed code image to the reference image (expected code image). Experiments show that the code image can survive various graphical transfer processes, such as halftone screening, printing and digitizing. An actual document containing a printed Full Spectrum structure, the Security ID, is used as a practical example.