A low-complexity reduced-reference print identification algorithm

In a print production system, the ability to match a printed document with its original electronic form enables services that improve robustness of the production process, such as content tracking mechanisms, or highly targeted quality assurance checks. One approach to this problem is to use overt markings that are later removed. This work, however, adopts a method that uses no markings, relying instead on fast image matching criteria. Though general image matching can be a difficult problem, constraints associated with the production environment allow construction of a simple, efficient solution to the matching problem. A reduced-reference quality assessment algorithm is presented that is used to match every document in a collection of printed and scanned documents with its original electronic form, based on down-sampling, equalization, quantization and per-pixel comparison. Using a 1085 image test set, a 100% matching rate is achievable by representing each image with only 5 bits-per-pixel at 8 dots-per-inch. The proposed method also compares favorably with other current reduced-reference quality assessment algorithms in terms of speed, storage and achieved matching rate.