Searching for discriminating features in B&W inkjet prints to individualize printers in a forensic setting

Abstract

In recent years several authors have demonstrated that it is possible to identify a certain printer model by the use of image texture analysis. However, these articles often describe only a limited number of printers (typical 10) and non realistic test targets.In this study we search for discriminating features in B&W inkjet prints, not only to be able to exclude a specific printer as a possible source, but also for individualizing printers. The scope of the study is anonymous letters and forged (financial) documents. The study is performed using a realistic test document, which only incorporates text and symbols. This document is printed on the same printing paper on more than 75 different printers and on three sets of 10 identical printers, four samples on each printer. The samples will be analyzed with the use of the ImageXpert Full Motion system at more than 500 different locations on the document, measuring more than 50 different features, such as line spacing, edge quality, line thickness, area and overspray.Advanced data analysis will be performed on the measured data to select the best (combination of) discriminating features.This article will describe the scope and goal of the study, the design of the test document and the research strategy. Furthermore, the first preliminary results will be shown.