In the field of forensic science the question of finding out a solution to discriminate ink jet printings, provides interesting indicators to investigators. Nevertheless, this task is not obvious because of several parameters such as the media type, the artefact of the printer, the age of the printer and so on. The aim of this study is to identify automatically type and model of a printer from the characters of an anonymous letter for instance. Generally, it is not possible to distinguish a printer from another, just by a visual inspection of the writting. Our work based on recognition pattern consists in finding some features extracted from the letter « a », able to characterize the printer. A stochastic approach is used to identify the invariant features of the letter « a » printed by three kinds of printers: Epson Stylus, Canon i905, HP Photosmart. The principle of the method is based on the calculation of seven invariant moments proposed by Hu3. The distribution of each moment is modelled by a gaussian component in a training phase which contains 80 letters. The test phase is based on different conditions. The first one consists in identifying a printer. The second one evaluates the influence of three word processing software, and at last, the third one proposes a study of the scanner effect. The obtained results reveal that printer discrimination is possible independently from the word processing software. And last the scanner effect decreases significantly the power of discrimination according to the resolution used.
Vanessa Talbot, Patrick Perrot, Cyril Murie, "Inkjet Printing discrimination based on invariant moments" in Proc. IS&T Int'l Conf. on Digital Printing Technologies (NIP22), 2006, pp 427 - 431, https://doi.org/10.2352/ISSN.2169-4451.2006.22.1.art00029_2