Académique Documents
Professionnel Documents
Culture Documents
AbstractIn
watermark may be difficult to hide. On the other hand, if the watermark is embedded in perceptually insignificant components, it would be easier to hide the watermark but the scheme may be less to resilient to attacks [3].Most Researches pay attention to the development of watermarking schemes for grayscale images than color images [4]. In the proposed method, cover image is converted into RGB color spaces to decorrelate it then watermark of size 64x64 is hidden into any one color spaces of a cover image, obtained results are tabulated. Similarly input image is converted into YIQ color space and watermark is hidden into either in luminance or in chrominance components. Finally results of RGB color space is compared with YIQ space. In section I introduction about the necessity of watermarking is given. Section II discuss about preliminaries of transformation techniques used and SVD technique briefly. In section III proposed algorithm is given in detail as step by step procedure. Section IV discusses in detail the performance evaluation of proposed algorithm. Section V discusses about watermarking in YIQ color spaces. Based on results obtained the features of proposed algorithm is concluded in section VI.
II. PRELIMINARIES OF DWT-DCT-SVD
I. INTRODUCTION With the growth of networked multimedia systems, the need of secure communication and data transfer is increasing [1].Among todays information security techniques, multimedia authentication techniques become a powerful tool for protecting multimedia content. Watermarking is one of the multimedia authentication techniques [2].Watermarking is the process of embedding data into a multimedia element such as an image, audio or video file for the purpose of authentication. This embedded data can later be extracted from, or detected in, the multimedia for security purposes. A watermarking algorithm consists of the watermark structure, an embedding algorithm, and an extraction or detection algorithm. Watermarks can be embedded in the pixel domain or a transform domain. In multimedia applications, embedded watermarks should be invisible, robust and have a high capacity. The most important uses of watermarks include copyright protection and disabling unauthorized access to content [3]. In recent years, many watermarking schemes have been developed using DWT and DCT. If the watermark is embedded perceptually most significant components, the scheme would be robust to attacks but the
V.Santhi is with VIT University, Vellore, India (Cell: 9865762713; email: vsanthi@ vit.ac.in). N.Rekha and S.Tharini is with VIT University, School of Computing Sciences,Vellore, India,
A. Discrete wavelet transformation DWT is a new technique which is used to represent image in a new time and frequency scale in recent years. The basic function of DWT is decomposing the input signal to multiresolutions. If input signal is an image DWT decompose it into low frequency (LL) and high frequency (HL, LH and HH).HL represents the horizontal detail, LH represents the vertical detail and HH represents the diagonal part. The low frequency band is the optimal approximation of the original image, which is determined by the DWT decomposition progressions that represents the maximum scale and distinguishing degree [5]. B. Discrete cosine transformation DCT function is a transformation technique which transforms the input image from spatial domain to frequency domain. In DCT-based watermarking, the DCT coefficients are modified to embed watermark data [6]. DCT is used in many standardized image, audio, and video compression methods. It has shown its superiority in
reduction of the redundancy of a wide range of signals [2]. In proposed method block DCT is used to increase the robustness. C. Singular value decomposition SVD is a mathematical tool used to analyze matrices. In SVD, a given matrix A is decomposed into three matrices such that, A=USVT where U and V are orthogonal matrices and UT U=I, VT V=I, I is an identity matrix. The diagonal entries of S are called the singular values of A, the columns of U are called the left singular vectors of A, and the columns of V are called the right singular vectors of A. This decomposition is known as the singular value decomposition (SVD) of A [7]. The three main properties of SVD from the view point of image processing applications [8][9]: 1. The singular values of an image have very good stability, that is, when a small perturbation is added to an image, its singular values do not change significantly. 2. Each Singular value specifies the luminance of an image layer while the corresponding pair of singular vectors specifies the geometry of the image. 3. Singular values represent intrinsic algebraic properties. SVD belongs to a kind of transforming domain technique, the principle is to embed the watermark into the SVD domain of original image, but more time has to be spent to process the whole image. To enhance the efficiency of watermarking algorithm block-SVD can be used. [10]
III. PROPOSED ALGORITHM
4. 5.
6. 7. 8.
9.
Apply SVD to all DCT transformed blocks, B ij = Uij Sij VijT . Extract the ij maximum singular value in S ij, the maximum singular value from every block can build a new matrix C, which is again decomposed by SVD, C=U S VT Apply SVD to the watermark, W = Uw Sw VwT . Modify the singular values of C matrix by using the singular values of watermark. S1=S+Sw Combine the modified singular values with the orthogonal matrices of C,S2=U S1 V T . ij <= C <=S2 Replace the maximum singular values ij by the modified singular values ij in each block and apply inverse SVD technique to each block B ij <=U ij S ij V ij T Apply inverse DCT to each block of low frequency band Repeat the same procedure for other bands of DWT Then perform IDWT to obtain the watermarked component. Combine the watermarked color component with other components to obtain the watermarked image, A*.
This algorithm combines the properties of DWT, DCT and SVD. DWT is used to decompose the original image in to four band of frequency. DCT technique is applied on each band to collect significant coefficients in which watermark is to be hidden. SVD technique is applied to the watermark and it is hidden into the host image by adding the singular values of host image and singular values of watermark. In this algorithm watermark is added to the SVs of whole image, which may lost due to attacks. To avoid this disadvantage, we propose an approach in which DWT coefficients are converted into 44 blocks. Each block is DCT transformed followed by SVD to get singular values of each block. Maximum Singular value is extracted from each block to form a new matrix called singular block. SVD is applied to the singular block to embed the watermark. Embedding and extraction of watermark is shown in Fig.1. Let A be the cover image. Convert the cover image into RGB color space. Extract any one of the color component. Apply watermark embedding algorithm on each color component.
6.
7.
8.
The above mentioned embedding and extraction algorithm is executed by hiding data in low frequency band. Similarly, watermark can be hidden into other frequency band also. Instead of hiding data in RGB color space, it can be extended to YIQ or YUV color space. Proposed algorithm is tested in YIQ color space also because in YIQ color model image content are highly decorrelated.
LH HH New matrix C= U ij
Watermark W
SVD (U, S, V)
SVD(Uij S ij Vij )
Sw S2 = U S1 V S1= S+Sw
= (S
- S )/
* *
W = Uw Sw
Vw
New matrix * * C = U ij
Watermark W
SVD(U ij S ij V ij)
Inverse DWT
*
ij
LL HL
LH
* *
LL DWT HL
LH
* *
HH
HH
IV.PERFORMANCE EVALUATION
In this proposed algorithm the Lena image of size 512x512 is taken as test image and the size of watermark is 64 x 64. Selected embedding intensity value is 0.20 for low frequency band and 0.15 for other frequency bands. The original image, watermark and watermarked is shown in Fig 2. Similarly watermarked image and extracted watermarks are shown in Fig.3.
RT
Fig.2 (a)Host image (b) watermark (c) Watermarked image
RT RT RT RT
Fig.3 Extracted watermark from all Frequency Band
The quality of the watermarked image can be measured either subjectively or objectively and it is observed that both subjective and objective quality of watermarked image is good. The PSNR is the objective criteria used to measure the quality of the watermarked image. The Equation for measuring PSNR and NC are shown in (1), (2) and (3) respectively. After embedding the watermark in the host image based on DWT-DCT-SVD, the quality of watermarked image is measured using peak signal to noise ratio (PSNR) value and it is observed that the value is 26.7912. If the watermarked image is not attacked by unknown users then calculated normalized correlation (NC) is 1.Normalized correlation is a factor which is used to measure the similarity of extracted watermark with original watermark. Robustness of watermarked image is tested with various attacks such as rotation , salt and pepper noise, Gaussian noise, image sharpening, histogram equalization, Gaussian blur, color contrast and resizing if images. Extracted watermark from green color space after different kinds of attacks experienced by a watermarked image is shown in Table 1 and Table 2. Watermark can be hidden in any one of the color spaces like red or green or blue of cover image. Results obtained from green color space are compared with the results of red and blue color spaces which are shown in Table 3 , Table 4 and Table 5.
Watermarked image
Extracted wm from LL
Extracted Wm from LH
Extracted Wm from HL
Extracted Wm From HH
Compression
0.9816
0.9807
0.9690
0.9952
Rotation
- 0.6820
0.8996
0.9396
0.9548
Resize
0.8123
0.9926
0.9926
0.9813
Sharpening
0.9950
0.9979
0.9953
0.9982
Extracted Wm from LH
Extracted Wm from HL
Extracted Wm from HH
0.9459
0.9661
0.9596
0.9888
Gaussian blur
0.9999
0.9972
0.9787
0.9492
Color contrast
0.3669
0.9349
0.9254
0.9430
Histogram equalization
0.9743
0.9934
0.9917
0.9945
Gaussian noise
0.9911
0.9822
0.9830
0.9815
TABLE.3 CALCULATED CORRELATION VALUE OF EXTRACTED WATERMARK FROM RED COLOR SPACE
ATTACKS Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Color contrast Resize Compression
Input image is converted into YIQ color model in which luminance and chrominance components of input image is separated. Watermark is hidden into both luminance and chrominance components and results are tabulated in Table6, Table 7 and Table 8. It is observed that proposed algorithm is robust to rotation attack in YIQ space than RGB space. Thus as per the observation YIQ model is better than RGB model for information hiding.
(1) (2)
NC =
TABLE.4 CALCULATED CORRELATION VALUE OF EXTRACTED WATERMARK FROM GREEN COLOR SPACE
(3)
ATTACKS Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Color contrast Resize Compression
ATTACKS Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Color contrast Resize Compression
TABLE.5 CALCULATED CORRELATION VALUE OF EXTRACTED WATERMARK FROM BLUE COLOR SPACE
ATTACKS ATTACKS Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Resize Compression LL 0.9942 0.99602 -0.4288 0.9970 0.8240 0.9999 0.9766 0.9740 LH 0.9674 0.9828 0.9007 0.9970 0.9893 0.9938 0.9841 0.9714 HL 0.9718 0.9855 0.9279 0.9945 0.9892 0.9359 0.9834 0.9617 HH 0.9618 0.9814 0.9549 0.9977 0.9908 -0.4488 0.9721 0.9831 Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Color contrast Resize Compression
REFERENCES
[1] F. Hartung and M. Kutter, "Multimedia Watermarking Techniques," in proc. of the IEEE, vol. 87, no. 7, pp. 10791107, July 1999. Feng Liu, Yangguang Liu,A watermarking algorithm for digital image based on DCT and SVD, Congress on Image and Signal Processing 2008.
ATTACKS Salt and peppers noise Gaussian noise Rotation Sharpening Histogram equalization Gaussian blur Color contrast Resize Compression
[2]
[3]
Navas K A, Ajay Mathews Cheriyan, Lekshmi. M, Archana Tampy.S, Sasikumar M, DWT-DCT-SVD Based Watermarking 3rd International Conference on Communication Systems Software and Middleware and Workshops, 2008. Kuo-cheng Liu and Chun-Hsien Chou, Robustness comparison of color image watermarking schemes in uniform and non-uniform color spaces, IJCSNS International Journal of Computer Science and Network Security, Vol No.7, July 2007. Cheng-qun, Li Li, An-Qiang Lv and Li Qu, Color Image Watermarking Algorithm Based on DWT-SVD, International conference on Automation and Logistics, August 18-21, 2007 Feng Liu, et al. A Watermarking Algorithm for Digital Image Based on DCT and SVD Congress on Image and Signal Processing, 2008. B.Chandra Mohan,S.Srinivas Kumar,A Robust watermarking scheme using singular value decomposition,Jounal of Multimedia, vol.3, No.1, May 2008. Liu Liang and Sun Qi,A new SVD-DWT composite watermarking,ICSP2006 proceedings Zhe-Ming Lu, Hong-Yi Zheng, Ji-Wu Huang, A Digital Watermarking Scheme Based on DCT and SVD.
[4]
VI CONCLUSION
1. 2. 3.
4.
5. 6.
In proposed algorithm the features of DWT-DCTSVD techniques are combined. Embedding can be carried out in all frequencies without visible distortion in cove image Algorithm is found to be robust against many attacks as per the results, classified as non-blind watermarking algorithm, which can be used for copyright protection application where robustness is very important. Watermark inserted in low frequencies are resistant to certain types of attacks, and watermarks embedded in high frequencies are resistant to certain type of attacks. Thus based on applications low, middle or high frequency band is selected for hiding watermark. Similarity measure shows that our proposed algorithm outperforms in YIQ color space rather than RGB color space. Proposed algorithm is not robust to rotation attack in RGB color space but robust in YIQ color spaces. Thus our future work is to make algorithm robust to rotation attack in RGB color space.
[5]
[6]
[7]
[8]
[9]
[10] Xinzhong Zhu, Jianmin Zhao and Huiying Xu,A Digital watermarking Algorithm and implementation based on improved SVD, The 18th International conference on pattern recognition, 2006.