Document recognition/authentication based on medium-embedded random patterns

An approach to authenticating/recognizing documents based on random patterns embedded into the document medium is described. The random patterns are formed by embedding short fibers during medium fabrication and the resulting documents have extremely high uniqueness. Therefore, these patterns offer an effective anti-counterfeiting measure. The proposed algorithm compares a document under consideration to all documents contained in a database assuming that the newly acquired document may be rotated and translated relative to its match in the database, and/or parts of the document may be missing. The algorithm has been tested on over 2000 synthetic images always giving satisfactory results