Document authentication

This thesis deals with document authentication. The test document is compared with documents in the database in order to determine a possible match. The database consists of a large number of different documents containing fibers randomly distributed over noisy backgrounds. The test document may be translated and rotated relative to documents in the database. Also, some fibers may be missing in the test document relative to those in the database. In this thesis a new approach to document authentication based on distances between fiber end points is introduced. The algorithm has been tested on more than a thousand different documents. The thesis analyzes the probability of uniqueness of documents represented by fibers as well as the probability of uniqueness of documents represented by fiber end points. To overcome difficulties caused by the large database, assigning a unique signature to documents is considered in the thesis, too. Three different ways of assigning signatures to documents are considered, and all of them are based on the concept of distances between fiber end points.