Feature Based Watermarking Algorithm by Adopting Arnold Transform S.S. Sujatha 1 and M. Mohamed Sathik 2 1 Assistant Professor in Computer Science, S.T. Hindu College, Nagercoil, Tamilnadu, India 2 Associate Professor in Computer Science, Sadakathullah Appa College, Tirunelveli, Tamilnadu, India sujaajai@gmail.com, mmdsadiq@gmail.com Abstract. The central idea of this paper is to develop an algorithm that embeds the watermark information in host image to authenticate it. The host image is divided into non-overlapping blocks of size 2x2. Minimum value is obtained from each block and the resultant matrix is scrambled for three times with Arnold transform to enhance security. From the transformed matrix a binary watermark is constructed and is embedded within the host image. The operation of embedding and extraction of watermark is done in high frequency domain of Discrete Wavelet Transform since small modifications in this domain are not perceived by human eyes. Furthermore, the proposed algorithm is checked against various common image processing attacks. This watermarking scheme deals with the extraction of the watermark information in the absence of original image, so the blind scheme was obtained. Keywords: Digital watermarking, Discrete Wavelet Transform, Arnold transform, Image Authentication, Content based watermarking. 1 Introduction Internet plays a very important role in accessing information from any place. The development of web technology makes it possible for us to transfer digital content in a cost-efficient way. However it causes serious problems such as unauthorized use and manipulation of digital content. Digital watermarking is a technique which embeds additional information called digital signature or watermark into the digital content in order to secure it. Several methods have proposed in literature. A survey is in [1]. The proposed methods can be classified according to the domain in which the embedding is performed. They are spatial-domain techniques and frequency-domain techniques. Least Significant Bit (LSB) is the simplest technique in the spatial domain [2] which directly modifies the intensities of some selected pixels. The frequency domain technique transforms an image into a set of frequency domain coefficients [3]. The transformation adopted may be discrete cosine transform (DCT), discrete Fourier transforms (DFT) and discrete wavelet transforms (DWT) etc. After applying transformation, watermark is embedded in the transformed coefficients of the image such that watermark is not visible. Finally, the watermarked image is obtained by acquiring inverse transformation of the coefficients. V. V Das, R. Vijaykumar et al. (Eds.): ICT 2010, CCIS 101, pp. 78 82, 2010. Springer-Verlag Berlin Heidelberg 2010
Feature Based Watermarking Algorithm by Adopting Arnold Transform 79 In the proposed content based watermarking scheme, watermark is generated from the feature of image. Discrete wavelet transform (DWT) is used for embedding watermarks, since it is an excellent time-frequency analysis method, which can be well adapted for extracting the information content of the image [4]. Watermark has been embedded in the high frequency bands. In this paper we propose a novel DWT transform based watermarking scheme which is robust against many common image attacks and experimental results verify this. 2 Research Preliminaries This section briefly describes the techniques and methods that have been adopted by the proposed scheme, including DWT and watermark scrambling using Arnold transform. 2.1 Discrete Wavelet Transform The proposed algorithm is a frequency domain watermarking scheme and works by modifying the DWT coefficients. The wavelet transform decomposes input image into four components namely LL, HL, LH and HH. The lowest resolution level LL consists of the approximation part of the original image. The remaining three resolution levels consist of the detail parts and give the vertical high (LH), horizontal high (HL) and high (HH) frequencies. In the proposed technique, embedding and extraction of watermark takes place in the high frequency component. For a one level decomposition, the discrete two-dimensional wavelet transform of the image function f(x, y) is found in [7]. 2.2 Arnold Transform A digital image can be considered as a two unit function f(x, y) in the plane Z. It can be represented as Z = f(x, y) where x, y { 0,1,2,3... N 1} and N represents order of digital image. The image matrix can be changed into a new matrix by the Arnold transform which results in a scrambled version to offer security. It is a mapping function which changes a point (x, y) to another point (x 1, y 1 ) by the equation (1). 3 Proposed Method 1 x 1 y = 1 1 1 2 x mod y The proposed watermarking scheme consists of three phases: Watermark generation, Watermark embedding and Watermark detection. A 1-level Discrete Wavelet Transform is performed for embedding watermark. Watermark information is embedded in the high frequency bands (HH1) since it is robust against various normal image processing and malicious attacks. The watermark is generated from the pixel value of original image and so there is no need of external image or logo. Hence a method is necessary to generate watermark. The procedure for generating watermark data is summarized below. N (1)
80 S.S. Sujatha and M. Mohamed Sathik 3.1 Watermark Generation The watermark is generated from the spatial domain information. It includes the following steps: The original image X of size M x N is partitioned into non-overlapping blocks of size 2 x 2. Compute minimum value from each block and construct a matrix M b (p, q),,2,3... / 2 1,2,3... N / 2. where p { 1 M } and q { } Find median value M d for the elements in M b (p, q). Perform Arnold transform for three times on M b (p, q) to scramble the elements and obtain matrix M s (p, q). Form the watermark pattern to be embedded into original image as 0 if M ( p, q) > M s d W ( p, q) = 1 otherwise For a M x N image, a watermark pattern of size M/2 x N/2 is generated. In this method, watermark is generated by performing some operations on image pixels rather than taking from external source, hence the name content based watermark. 3.2 Watermark Embedding The watermark is embedded in the high frequency subband of DWT as follows: Apply 1-level DWT to original image. The watermark is embedded in the high frequency component HH1 of DWT. Perform inverse wavelet transform (IDWT) to obtain the watermarked image. 3.3 Watermark Detection Proposed watermarking scheme deals with the extraction of the watermark information in the absence of the original image and so it can be referred as blind watermarking. The authentication process includes the following steps: Calculate watermark from watermarked image using the steps described under watermark generation in section 3.1. Apply 1-level DWT to watermarked image and extract watermark from HH1 subband. Compare the two watermarks (calculated and extracted) and decide the authenticity. 4 Experimental Results In this paper, we consider the images with number of rows and columns are exact multiples of 2 since the image is divided into blocks of size 2x2. For testing, the size
Feature Based Watermarking Algorithm by Adopting Arnold Transform 81 of the original image is taken as 512x512. Fig.1 (a) shows original image. A 256x256 watermark signal is constructed from original image and is embedded within itself. The proposed method is demonstrated with several images using MATLAB. After embedding the watermark, the watermarked image that is produced and the original image cannot be distinguished by naked eye. Fig.1 (b) shows watermarked image. The absolute difference of the pixel intensities of the watermarked image and the original image is shown in Fig.1(c). The visual quality of original and watermarked images is measured using the Peak Signal to Noise Ratio and a comparison between extracted and original watermark can be done by computing Similarity Ratio (SR) between these two patterns as defined in [10]. (a) (b) (c) Fig. 1. (a) Original image, (b) Watermarked image, (c) Difference image The proposed algorithm was tested with several attacks. Table 1 gives the performance of proposed watermarking scheme under various attacks. Adding Gaussian noise (mean, variance) Adding Salt & Pepper noise Median filtering Linear filtering Image adjustment Histogram Equalization Table 1. Quality evaluation of proposed scheme Attacks PSNR(dB) SR No 54.1047 1 0.001, 0 54.1047 1 0.005, 0 45.6804 1 0.01,0 37.8092 1 0.05,0 25.6831 0.9992 0, 0.001 30.0730 0.5188 0,0.002 27.1791 0.5123 0.001 35.1050 0.9898 0.002 32.0153 0.9861 0.005 28.1179 0.9751 0.01 25.0121 0.9496 3x3 29.5819 0.5218 3x3 27.7792 0.5359 18.7003 0.8433 19.0192 0.7573
82 S.S. Sujatha and M. Mohamed Sathik The experimental results show that the qualities of watermarked images are satisfactory in the case of filtering and little additive noises since the PSNR values are greater than 25. The SR values are all greater than 0.76 (except filtering and some cases of Gaussian noise) indicates that the watermark is robust to common image processing attacks. 5 Conclusion This study has proposed a robust watermarking technique based on the frequency domain. In the designed method watermark is constructed from the image by applying Arnold transform to enhance security of it. It also makes use of the Discrete Wavelet Transform which provides a frequency spread of the watermark within the host image. Moreover the authentication process is simple, fast and allows easy extraction. The performance of the watermarking scheme is evaluated with some possible image processing attacks. Experimental results revealed a good resilience against such attacks. However the proposed method is not maintaining quality of watermarked images against attacks such as Histogram Equalization and Image adjustment. Future work may be concentrated on improving quality and resistance against those attacks. References 1. Rey, C., Dugelay, J.: A Survey of Watermarking Algorithm for Image Authentication. Journal on Applied Signal Processing 6, 613 621 (2002) 2. Podilchuk, C.I., Delp, E.J.: Digital Watermarking: Algorithms and Applications. IEEE Signal Processing Magazine, 33 46 (2001) 3. Parthasarathy, A.K., Kak, S.: An Improved Method of Content Based Image Watermarking. IEEE Transaction on Broadcasting 53(2), 468 479 (2007) 4. Reddy, R., Prasad, M.V.N., Sreenivasa Rao, D.: Roubst Digital Watermarking of Color Images under Noise Attacks. International Journal of Recent Trends in Engineering 1(1) (2009) 5. Xia, X.-G., Boncelet, C.G., Gonzalo: Wavelet Transform based Watermark for Digital images. Optics Express 3(12), 497 511 (1998) 6. Gonzalez, R.C., Woods, R.E., Eddins, S.L.: Digital Image Processing Using MATLAB, India (2008) 7. Kumar, S., Raman, B., Thakur, M.: Real Coded Genetic Algorithm based Stereo image Watermarking. IJSDIA 1(1), 23 33 (2009) 8. Liu, H., Rao, J., Yao, X.: Feature Based Watermarking Scheme for Image Authentication. IEEE, 229 232 (2008) 9. Dittmann, J.: Content-fragile Watermarking for Image Authentication. In: Proc. of SPIE, Security and Watermarking of Multimedia Contents III, vol. 4314, pp. 175 184 (2001) 10. Yuan, Y., Huang, D., Liu, D.: An Integer Wavelet Based Multiple Logo-watermarking Scheme. IEEE 2, 175 179 (2006)