Towards a feature set for robust printing-imaging cycle device identification using structured printed markings

Device identification is an emerging field where technologies used to create a digital image are inferred by strategic image analysis. Some of the more well understood topics in this area include techniques to identify cameras, scanners and printers. The goal of printing-imaging cycle device identification is to gather information about the printing and imaging devices used to create, then digitally reacquire, a physical document. It has been shown that accurate printer-scanner identification is possible using a specific type of marking composed of oriented, colored tiles. This paper discusses a set of features based on histogram analysis that can be applied to printing-imaging device identification problems using a wider range of small structured markings (color tile deterrents and Guilloche curve patterns). The features are robust to a number of different effects, including changes in perspective and scan resolution. The approach results in superior identification performance in comparison to other state-of-the-art strategies, and more robustness to marking orientation. Performance as a function of scan resolution is also discussed in detail.