COMPARISON OF SVD BASED IMAGE WATERMARKING TECHNIQUES IMPLEMENTED FOR GRAY SCALE AND COLOR IMAGES

VIJAYA KUMARI V.¹*, CHITRA B.^2
1Department of Electronics and Communication Engineering, Sri Krishna College of Technology, Coimbatore-641042, India
²Department of Electronics and Communication Engineering, Sri Krishna College of Technology, Coimbatore-641042, India
* Corresponding Author : ebinviji@rediffmail.com

Received : 06-11-2011     Accepted : 09-12-2011     Published : 12-12-2011
Volume : 3     Issue : 4       Pages : 359 - 363
Int J Mach Intell 3.4 (2011):359-363
DOI : http://dx.doi.org/10.9735/0975-2927.3.4.359-363

Conflict of Interest : None declared

Data authentication and data security are the primary requirement in present day communication system. In image processing, data authentication is implemented using watermarking technique. Recently several watermarking techniques have been proposed. The drawback of these techniques is less robustness, less fidelity and degradation in image quality. Watermarking applications are used in copyright protection, broadcast monitoring, usage control, and this work is implemented for copyright protection using singular value decomposition (SVD) technique. In this technique, the singular values of watermark image are embedded into singular values of host image to get watermarked image. Singular value decomposition (SVD based watermarking techniques is implemented for gray scale and color image of different sizes and the comparative analysis is made based on the results produced when the image is subjected to different types of attacks.

image processing, data authentication, decomposition, watermarking, compression, scaling, cropping.

Rapid development of digital multimedia technique led to easy copying and manipulation of digital data. Digital watermarking is an excellent tool used for copyright protection by embedding some information into digital data. This embedded watermark should be invisible, robust and should have high capacity. This embedded data can be later extracted to make an assertion about data authenticity. There are different types of watermarking based on the domain used for embedding watermark into the host image.
The singular values of watermark are multiplied by a scaling factor and added to singular values of host image [1] . The host image is divided into blocks and watermark is embedded into singular values of each block of host image [2] and this method is robust to JPEG compression, Gaussian noise and cropping. Host image is partitioned into four sub images, and the watermark is embedded into both D and U component of two sub images [3] and [4] . RDWT is applied to both cover and watermark images and SVD is applied to LL sub bands of both the images, by this method large watermark can be embedded into cover image [5] . Feature points are extracted from host image and several circular patches centered on these points are generated. These patches are used as carrier of watermark information [6] . SVD is implemented for color image watermarking [7] , where V component of host image is used for embedding the watermark. Hybrid method involving SVD and contourlet transform [8] is also implemented for color image watermarking.

Singular value decomposition (SVD) based watermarking technique, where singular values of host image are used as the carrier of watermark information is implemented on three different types of images as follows:
A) Gray scale image: host and watermark image of same size.
B) Color image (RGB): host and watermark image of same size.
C) Color image (RGB): watermark image (30X30) is smaller than host image (100X100).
These different types of watermarked images are subjected to different types of attacks such as addition of noises, blurring, sharpening, filtering etc., and comparative analysis is made based on the results produced for different types of attacks. The visual quality of watermarked image is measured using PSNR values using formula
PSNR = 10log [ (255^2) / (∑ ( I’(i,j) - I(i,j)) ^2 ] ………….. (1)
Where I’(i,j) is the watermarked image and I(I,j) is the original image.
Normalized cross correlation value is used for measuring the identity between original watermark and extracted watermark.
NCC = [ ∑∑W(x,y)*W’(x,y) ] / [ ∑∑(W(x,y))^2 ] …………… (2)
Where W(x,y) is the original watermark and W’(x,y) is the extracted watermark.

Step1: Host and watermark image is decomposed into separate R, G, B color spaces.
Step2: singular value decomposition (SVD) is computed for R color space of host and watermark image.
Step3: Modify the singular values of R color space of host image with singular values of R color space of watermark image using spread spectrum technique as follows.
λi' = λi + αλw ………………………………………………… (3)
Where α is scaling factor, λi is a singular value of host image, λw is a singular value of watermark image, λi' is a singular value of watermarked image.
Step4: Apply inverse singular value decomposition (SVD) on modified singular values obtained in step 3 to get the coefficients of watermarked image.
Step 5: Apply the steps 2, 3 and 4 for G and B color spaces.
Step 6: Combine the R, G, B color spaces of watermarked image to obtain the color watermarked image.

Step 1: Separate the R, G, B color spaces of water marked image.
Step 2: Apply singular value decomposition (SVD) to R color space of transformed watermarked image.
Step 3: Extract the singular values from R color space of watermarked and host image,
λw = (λi’- λi) / α …………………… (4)
Step 4: Apply inverse SVD to obtain the coefficient of R color space of transformed watermark image.
Step 5: Apply steps 2, 3 and 4 for G and B color spaces.
Step 6: Combine R,G and B color spaces to get the color watermark.

Host image (1a) and watermark image (1b) of size 256X256 are taken for this technique. Watermark image is embedded into host image and the watermarked image is shown in [Fig-1] (c). the extracted watermark without any attack is shown in [Fig-1] (d). The scaling factor of 0.3 is employed for all types of images implementing the embedding algorithm. The watermarked and extracted images for different attacks are shown in [Fig-2] and [Fig-3] respectively.

Host image (4a) and watermark image (4b) of size 200X150 are taken for this technique. Watermark image is embedded into host image using singular value decomposition (SVD) embedding algorithm and the watermarked image is shown in [Fig-4] (c). the extracted watermark without any attack is shown in [Fig-4] (d). The experiment is performed by taking scaling factor of 0.3. The watermarked and extracted images for different attacks are shown in [Fig-5] and [Fig-6] respectively.

Host image (7a) of size 100X100 and watermark image (7b) of size 30X30 is taken. Watermark image is embedded into host image using singular value decomposition (SVD) embedding algorithm and the watermarked image is shown in [Fig-7] (c). the extracted watermark without any attack is shown in [Fig-7] (d). The watermarked and extracted images for different attacks are shown in [Fig-8] and [Fig-9] respectively.
The results provided in Table1and 2 depicts that singular value decomposition (SVD) based watermarking technique provides good result for embedding smaller color image into larger color image, and not robust when implemented in gray scale image.

Singular value decomposition (SVD) based watermarking technique is implemented and comparative analysis is made for both gray scale and color images of different sizes, and their performance is evaluated using PSNR and NCC values. The Normalized correlation value obtained between original and extracted watermark image is unity for all types of images. Gray scale watermarking is not robust if noises are present and embedding smaller image into larger image produces better result in color image watermarking. The PSNR and NCC values are calculated for watermarked images without and with attacks. Further this can be extended for different types of attacks such as compression, scaling, cropping etc.

[1] liu R., Tan T. (2002) IEEE Trans. On multimedia, vol. 4.no 1, pp. 121-128.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[2] Ghazy R.A., El-Fishawy N.A., Hadhouds M.I., Dessouky M.I. and Abd El-Samie F.E., Ubiquitous Computing and Communication Journal, pp. 1-9.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[3] Chandra mohan D., Srinivas Kumar S. (2008) Journal of multimedia.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[4] Chin-Chen-Cheng Piyu Tsai & Chia-Chen Lin, (2005) Pattern Recognition Letters 26, 1577-1586.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[5] Samina Lagzian, Mohsen Soryani, Mohamood Fathy (2011) International Journal of Intelligent Information Processing, vol. 2, no. 1.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[6] Say Wei Foo, Qi Dong (2010) International Journal of Electrical and Computer Engineering, 5:6, 375-381.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[7] Rashmi Agarwal, Venugopalan K. (2011) IJCA special issue on, NCCSE-2011.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

[8] Venkata Narasimhulu C., Satya Prasad K. (2011) International journal of computer application, vo. 20, no. 8.  
» CrossRef   » Google Scholar   » PubMed   » DOAJ   » CAS   » Scopus  

	Fig. 1- Watermarked and extracted output without attack (a) Original image (b) Watermark mage (c) Watermarked Image PSNR = 39.09 d) Extracted watermark NCC = 1.00
	Fig. 2- Watermarked images for different attacks a) Sharpening b) Filtering c) Blurring
	Fig. 3- Extracted watermark images for different attacks a) Sharpening b) Filtering c) Blurring
	Fig. 4- Watermarked and extracted images without attacks (a) Original image (b) Watermark image (c) Watermarked image (Psnr = 37.466) (d) Extracted watermark NCC = 1.00
	Fig. 5- Watermarked images with different attacks (a) Poisson noise (b) Speckle noise c) Gaussian noise (d) Salt pepper noise (e) Filtering (f) Histogram equalization (h) Sharpening
	Fig. 6- Extracted watermark images for different attacks (a) Poisson noise (b) Speckle noise (c) Gaussian noise (d) Salt and pepper noise (e) Filtering (f) Histogram equalization (g) Blurring (h) Sharpening
	Fig. 7- Watermarked and extracted images without attacks (a) Original image (b) Watermark image (c) Watermarked image (psnr = 41.63) (d) Extracted image NCC = 1.00
	Fig. 8- Watermarked images with different attacks (a) Filtering (b) Gaussian (c) Salt and pepper noise (d) Histogram equalization (e) Blurring (f) Sharpening (g)Poisson noise (h) speckle noise
	Fig. 9- Extracted watermark images for different attacks (a) filtering (b) Gaussian (c) Salt and pepper noise (d) Histogram Equalization (e) Blurring (f) Sharpening (g) Poisson noise (h) speckle noise
	Table 1- PSNR values of watermarked images subjected to different attacks
	Table 2- NCC values of extracted watermark for different types of attacks