Development of an Electrophotographic Laser Intensity Modulation Model for Extrinsic Signature Embedding

In our previous work, we have demonstrated techniques to embed and extract extrinsic signatures from halftone images and text documents. Well developed embedding algorithms should increase the payload capacity while enhance the reliability of detection. In this study, we will develop a printer model that will be used to optimize the embedding algorithm for capacity and detection reliability. The model incorporates the impact of the process modulation parameter, e.g. laser intensity, with a stochastic dot interaction model to estimate the impact of the modulation on a known halftone pattern. Experimental data validated the effectiveness of the proposed model in predicting the impact of laser intensity modulation on the reflectance of the printout. Introduction Printer identification based on a printed document can provide forensic information to protect copyright and verify authenticity. In our previous work [1,2], we have demonstrated the feasibility of modulating dot gains through laser intensity modulation for halftone images and text document to embed unperceivable code sequences. To optimize the embedding and detections algorithms, significant amount of printing and measurements are needed. To reduce the time and effort, a suitable electrophotographic (EP) process model that characterizes the impact of laser intensity modulation on the printed image will be needed. In particular, a computation efficient model that characterizes the dot interactions among the printed pixels in a halftone pattern is a key component. Several dot interaction models for the EP process have been proposed in the literature. Roetling and Holladay [3] proposed the hard circular dot (HCD) model to account for spreading of colorant on the substrate that causes increased absorptance of the print. In this model, each printer addressable dot is assumed to be a circular spot with constant absorptance, and dots overlap are resolved by a logical OR at each point in the print. Pappas et al. [4] parameterized the HCD model and proposed a method to obtain the printer model parameters through macroscopic measurements of the test patterns [5]. Rosenberg [6] used HCD model to predict the tone response of various halftone algorithms and compensates the monotonic printer distortions before halftoning. Baqai and Allebach [7] incorporated the HCD model into the direct binary search (DBS) halftone algorithm to minimize the perceived meansquare error between the halftoned image and the continuous-tone image. To account for the dot scatter in EP process, Lin and Wiseman [8] proposed a stochastic dot model to model the toner particles distribution on addressable dots. Lin [9] used this model to improve the pattern uniformity and tonal response of frequency modulated halftone screens. Flohr et al. [10] also used a similar stochastic dot model to improve the halftone image quality produced by DBS. In this work, we will modify Lin’s [8] stochastic model to include the effect of laser intensity modulation used for extrinsic signature embedding to predict the reflectance of the printout. Using the proposed model, modulation threshold of various embedding algorithms can be efficiently estimated with fewer measurements. In addition, the modulation signal can be optimized to enhance the capacity and detection reliability through feedback of the estimated values. In this study, an HP Color Laserjet 4500 is used as the experimental platform. The remainder of this paper is organized as follows. Characterization of laser intensity modulation is discussed in the next section followed by the development of the stochastic dot interaction model with laser intensity modulation. Preliminary experimental results will be presented. Conclusions will be summarized in the last section. Characterization of Laser Intensity Modulation The intensity profile of the laser is modeled as a 2-D Gaussian envelope given by. 2 2 2 0 2 2 ( , , ) ( ) exp 2 2 y x I x y t I t W m α β ⎛ ⎞ ⎡ ⎤ = ⋅ − − ⎜ ⎟ ⎣ ⎦ ⎜ ⎟