Multipurpose Image Watermarking Scheme based on Self-embedding and Data Hiding into Halftone image Carlos Santiago Avila Section of Posgrate Studies and Research, ESIME CULHUACAN-IPN Mexico City, Mexico e-mail: csantiagoa0900@ipn.mx Mariko Nakano Miyatake Section of Posgrate Studies and Research, ESIME CULHUACAN-IPN Mexico City, Mexico e-mail: mnakano@ipn.mx Abstract—Nowadays the digital problems, such as copyright violation and content alteration, have been grown. To solve these problems, many watermarking based schemes have been proposed, however most of them address to solve one of these two problems. This work shows a way to embed the watermark for both purposes: copyright protection and content authentication. The main scheme is based on a self- embedding algorithm which generates the watermark sequence from the original image using halftoning method. The halftone watermark image furthermore is watermarked modifying their pixels to embed logotype related to the owner’s copyright; and the watermarked halftone image is embedded for content authentication using Quantization Index Modulation (QIM) in the Integer Wavelet Transform (IWT) domain. I. INTRODUCTION The growing demand to reduce the storage space and the easy exchange of digital information causes some serious problems, such as copyright violation and content alteration. For example, in July 2005 it was discovered that a number of Second World War files held at the National Archives contained forged documents. An internal investigation found that the forgery took place during or after the year 2000 [1]. Initially, watermarking techniques have been employed for copyright protection of digital materials, such as image, audio and video [2]. And later fragile and semi-fragile watermarking techniques were introduced for authentication purpose [3,4]. Recently multipurpose watermarking technique, in which both purpose: copyright protection and authentication are accomplished at same time, have attracted attention [5,6]. The multipurpose watermarking algorithm proposed Lu et al. [5] uses the relationship between the DWT coefficients of subbands with different decomposition levels and orientations. In [6], authors proposed multipurpose watermarking scheme using multiscale Curvelet Transform. Both schemes have tamper detection capabilities; however tampered regions cannot be recovered. In this paper, we propose a multipurpose watermarking scheme in which the watermark sequence is a halftone version of the original image. The halftone image is generated using Floyd-Steinberg error diffusion method, and then it is embedded into the original image using Quantization Index Modulation (QIM) method in the Integer Wavelet Transform (IWT) domain for content authentication purpose. Before the watermark embedding, the halftone image is furthermore modified to embed a binary logotype related to the owner’s copyright using the algorithm proposed in [7]. This algorithm was proposed to authenticate binary images, included scanned documents images [7]. To evaluate the tamper detection capability and robustness of the watermark sequence for copyright protection against JPEG compression, computer simulations are carried out. The experimental results show the desirable performance of the proposed scheme. The remainder of this paper is organized as follows. In Section 2, the proposed scheme is described and evaluation results are shown in Section 3. Finally in Section 4 some conclusions are remarked. II. PROPOSED MULTIPURPOSE WATERMARKING SCHEME IWT was chosen in place of the Discrete Cosine Transform (DCT) as embedding domain, because in the IWT the Human Visual System (HVS) is better related. The IWT provides a decomposition of an image into four sub- bands with integer elements, which are the approximation, and the three details with different orientations (horizontal, vertical and diagonal) by means of a convolution-based algorithm using high and low pass filters. Once the sub- bands (LL, LH, HL, HH) were obtained, the sub-band LL was chosen to embed the watermark sequence, because it is most robust although it cause more distortion. The watermark sequence was obtained from the original image using the Floyd-Steinberg error diffusion method [8] to convert the original image into binary halftone image, and then it is embedded into LL sub-band. Figure 1 shows the original gray scale image and halftone image generated by the Floyd-Steinberg method. Before the watermark embedding, the watermark sequence (halftone image of the original image) is modified using the method reported by [7] for copyright protection purpose. This embedding scheme is describe in the following section. 2010 Electronics, Robotics and Automotive Mechanics Conference 978-0-7695-4204-1/10 $26.00 © 2010 IEEE DOI 10.1109/CERMA.2010.112 394