International Journal of Computer Applications (0975 – 8887) Volume 128 – No.15, October 2015 13 A Print-Scan Resilient Watermarking based on Fourier Transform and Image Restoration R. Riad IRF-SIC, Ibn Zohr University, Agadir, Morocco H. Douzi IRF-SIC, Ibn Zohr University, Agadir, Morocco M. El Hajji IRF-SIC, Ibn Zohr University, Agadir, Morocco R. Harba PRISME, University of Orleans, Orléans, France F. Ros PRISME, University of Orleans, Orléans, France ABSTRACT The print-scan operation is still challenging in the watermarking community, some watermarking techniques were proposed in the literature to deal with this operation. These watermarking techniques are still very sensitive to degradations produced by the print-scan process. This paper investigates a watermarking technique in the Fourier domain that is robust to degradation produced when an image is printed in physical support then rescanned. The watermark is embedded in a middle frequency band of the discrete Fourier transform of an image using the improved spread spectrum technique. Some image restoration techniques were implemented. They were tested on images which were watermarked, printed and then rescanned before the watermark extraction. Experimental results clearly show the advantage of using the proposed approach. Keywords Watermarking, Fourier transformation, spread spectrum, print-scan, image restoration. 1. INTRODUCTION Image watermarking may be used to verify the authenticity of identity documents where pictures are present. In [1], a watermarking process for plastic card supports was proposed. The image is first watermarked, and then printed on the plastic support. To check if the watermark is present in the image, the document must be scanned (see figure 1 for more details).The so-called print-san operation can strongly reduce the efficiency of the watermark extraction. The watermarks should be robust enough to survive to various attacks produced in the print-scan process. At the same time, the embedded watermark should not degrade the visual quality of the image. The essential requirements of digital watermarking are robustness, perceptual transparency and capacity. In addition, watermark embedding and retrievals should have low complexity and be real time in order to be acceptable for various industrial applications. Adapted strategies must be developed in that context. The print-scan operation leads to a complex combination of different attacks, which produces various distortions. In [2], Lin and Chang separated these distortions into two main categories: geometric distortions and pixel value distortions. In the first case, those distortions consist in rotation, scaling and translation. In the second case, distortions are caused by luminance and color variations, contrast modifications, gamma correction, and blurring. Several approaches that counterattack the general geometric distortions have been developed in the literature; such as invariant transform [3], template insertion [1], feature-based algorithms [4], autocorrelation based method [5], and circularly symmetric watermark embedding in the DFT domain [6]. For the pixel value distortions, experimental models of the print-scan channel were proposed in [7]. For simplicity, the print-scan attack can be modeled as a low pass filter plus an additive noise independent of the image. A natural idea to reduce pixel value distortions could be to use a deconvolution method before the watermark extraction or some enhancement filters [8]. Fig 1: A possible scheme for securing smart card, insertion (top) and detection (below) of information in a smart card. In this paper, the watermark is embedded in the Fourier domain using spread spectrum technique. Before the watermark extraction a preprocessing step based on image enhancement techniques were applied the print/rescanned images. This preprocessing has as goal to reduce image blurring that occurs during the print-scan process. The paper is organized as follows: section 2 describes the chosen watermarking method. Section 3 presents the pretreatment method based on image deconvolution. Section 4 shows the experimental results. The conclusion and future works are described in the final section.