IEEE 2nd International Conference on Communication, Control and Intelligent Systems (CCIS)
A Robust image Watermarking Technique using 2-level Discrete Wavelet Transform (DWT) Rita Choudhary Dept. of Electronics, Rajasthan Technical University, Kota, Rajasthan, India E-mail Id:
[email protected] Abstract—The exponential growth in digital data over the internet has increased the requirement of a robust and high quality watermarking techniques. In general, the image watermarking techniques embed the binary or grayscale watermark into the cover image or into many multimedia images. In this method, variable visibility factor is used for the insertion of watermark into the low frequency component of the host image. In this paper, DWT-based image watermarking is proposed using level i.e. 2-level and also its parameters such as PSNR and NCC are compared with respect to 1-level DWT. The invisibility of watermarks generated using proposed method is depicted in the simulated results. Keywords: Image Watermarking, 2-level DWT, Wavelet Transform, PSNR
Girish Parmar Dept. of Electronics, Rajasthan Technical University, Kota, Rajasthan, India E-mail Id:
[email protected] Frequency domain is mainly used because of its robustness and invisibility which is the primary requirement of watermarking technique whereas spatial domain based watermarking easily affected by attacks and have the capacity problem. Therefore, the frequency domain based watermarking is preferred [5]. There are various types of transforms used in the frequency domain as shown in Fig. (1). Least Significant Bit Spatial Domain Watermarking
I. INTRODUCTION In this digital world, there is rapidly growing and sharing of multimedia files. Billions of digital files are transferred via the internet per day. There are many advantages privileged to users [1]. But there are numerous disadvantages also like a security problem, copyright of data, data transformation. The main source of these problems is internet. So security of the authentic information and some other issues has become a big question with multimedia source and content. Digital data can be easily copied and transferred to another user without loss of data and quality of data. Solving these types of problems are challenges for the researchers [2]. Data hiding technique is the solution of these problems. Some multimedia data is introduced into the owner’s data; further to prove the owner’s data extracted process is used. This idea was initially used in bank currency notes. The watermark is embedded in the currency notes to hide the originality of notes. Digital watermarking is a technique in which the original data is hidden with the watermark to preserve the originality of the original data. There are many requirements of watermarking techniques i.e., copyright protection, robustness, invisibility, data payload etc. Digital Watermarking technique can be classified by many different ways i.e. according to the domain, the type of document, human perception, the application. Mainly watermarking can be classified on the basis of working domain i.e. spatial domain and frequency domain [6].
978-1-5090-3210-5/16/$31.00 © 2016 IEEE
Patch Work On the basis of Domain Discrete Cosine Transformation Frequency Domain Watermarking
Discrete Wavelet Transformation
Fig. 1: Types of Watermarking on the Basis of Domain
DWT is a technique that collects the information about the frequency and location (in time), that’s why DWT can be preferred as compare to other wavelet transforms [3]. Further performance of image watermarking using DWT can be improved by increasing the levels of DWT i.e. 2-level DWT, 3-level DWT and so on. II. DISCRETE WAVELET TRANSFORMS (DWT) It is a technique for separating an image into various different sub bands i.e. •
Low-low (LL) sub band,
•
Low-high (LH) sub band,
•
High-low (HL) sub bands and
•
High-high (HH) sub bands.
Numerical analysis and functional analysis can be done by using different transformation techniques; they transformed data into another form but not the shape. It is useful for analyzing signals of a non-stationary nature [7]. As shown in Fig. (3) an analog filter is used for DWT. Four sub-bands LL1, LH1, HL1 and HH1 are obtained as output of these filters. Generally, the most of the energy is concentrated at the lower resolution part of the image. Therefore, the watermark is embedded in the lower part of the cover image and resulted watermarked image is more robust without losing the quality of the image [9-10]. On the other hand, the high frequency sub bands are not preferred for watermarking.
decomposition and other bands (HH, HL and LH) gives the smaller value of coefficient. LL band is more significant because it has maximum magnitude of the wavelet coefficient. The lower resolutions are computationally more effective for watermark detection because at every successive resolution level there are few frequency bands involved. DWT uses wavelet filters for image transformations. There are many types of filters that are used in transformation. The filters used commonly for watermarking technique are:•
Haar Wavelet Filter,
•
Daubechies Orthogonal Filters,
•
Daubechies Bi-Orthogonal Filters,
•
Symlet wavelets
•
Morlet wavelets
•
Coiflets wavelets
•
Meyer wavelets etc. III. WATERMARKING TECHNIQUE
Here, we introduced watermarking technique which has been implemented using 2-level DWT. The planned algorithm is separated into two steps, watermark embedding procedure and watermark extraction procedure. A. Embedding Procedure Here in embedding method, initially we read the host image and applied 1-level DWT on that image; this type of transform method decomposed it into different frequency components i.e. LL1, LH1, HL1 and HH1. Again 1-level DWT is applied on LL1 component of the host image. Now, we also take watermark image and applied the 2-level DWT on it. Now for embedding procedure LL sub band of the host (cover) image and watermark image are taken. Harr wavelets are used in this method. Now, add the LL sub band of the watermark image with the LL sub bands of the cover image after multiplied it by visibility factor (k). Even, we also use other sub bands of watermark image and cover image for embedding procedure. The watermarked image is obtained by using the formula
Fig. 2: DWT Decomposition Structure
Figure 2 illustrates the DWT decomposition structure. In the 1-level DWT the size of each part of the divided image is one-fourth of the original image. Fig. 3 represents the decomposition of the image using 2-level DWT. DWT is one of the most common methods used for image processing. This DWT transform method separates the image into four parts and the properties of human visual system exactly reflected by these wavelet coefficients.
WMLL2=LL2+k*LL2a;
(1)
Where WMLL2 = low frequency element of watermarked image, LL2 = low frequency element of the host image, LL2a =low frequency element of Watermark image, and k is visibility factors. Fig. 3: Decomposition of Image using 2-level DWT
A wavelet-based watermarking method is used to gain the robustness which also preserves both type of information of transformed data i.e. the frequency and the spatial information. Simple Filters can be used to implement wavelet transform efficiently and easily. Approximation coefficient (LL) band gives the larger magnitude of DWT coefficient at each level of 121
After embedding process of these two images, IDWT is applied to achieve the final watermarked image. B. Extraction Procedure In the extraction procedure, initially read the watermarked image and cover image and applied DWT which decomposed the images into different frequency elements.
To recover the watermarked image we use the formula:RW= (WMLL2-LL2)/k;
C. Simulated Experimental Results
(2)
Host Images
Where RW= Low frequency element of Recovered watermark, WM LL2= Low frequency element of the watermark image, and LL2= Low frequency element of watermarked image. After extraction process, to obtain the final extracted watermark image 2-level IDWT is applied. IV. RESULTS AND DISCUSSION
1. Lena
Here, a detail performance analysis of proposed technique has been presented. These experiments illustrate the efficiency of proposed watermarking technique.
2. Butterfly
Fig. 4(a): Host Image Each of (512*512) Size
Watermark Image
A. Materials and Methods Overall analysis has done with the 512X512 image and evaluated with considering fidelity parameters. Here, images used which are obtained from USC-SIPI image database as a standard evaluation database for watermarking algorithms. MATLAB software is used for simulation of this technique. B. Evaluation Fidelity Parameters
1. Dell Logo
The illustration performance of watermarked images is evaluated by using different fidelity factors i.e. peak signal-tonoise ratio (PSNR) and Normalized Correlation which are traditionally adopted in the field of image processing to calculate the performance of the output results as shown in tables. 𝑀𝑆𝐸 =
∑
∑
(𝑎(𝑖, 𝑘) − 𝑏(𝑖, 𝑘))
𝑃𝑆𝑁𝑅 = log
Fig. 4 (b): Watermark Image
The original images and watermark images used for this technique are depicted by Fig. 4(a) and Fig. 4(b) respectively. TABLE 1: SHOWS THE PSNR VALUE FOR D IFFERENT LEVEL OF DWT BASED IMAGE WATERMARKING T ECHNIQUES (LENA ( AS COVER) & DELL LOGO ( AS WATERMARK)
(3)
Visibility Factor α 0.04 0.03 0.025 0.02 0.01 0.005
(4)
Here, a and b are the two images at different locations (i,k). From eq. (4), L shows pixels range. The unit of PSNR is decibels (db). PSNR is depends upon mean squared error. Deviation of images measured using equation (4). The superiority of the image is calculated using normalized cross co-relation (NCC) and is obtained by using eq. (5) 𝑁𝐶𝐶 =
𝑸 ∗ ∑𝑷 𝒊 𝟏 ∑𝒌 𝟏 𝒃(𝒊,𝒌)∗𝒃 (𝒊,𝒌) 𝑸 𝟐 ∑𝑷 ∑𝑸 (𝒃∗ (𝒊,𝒌))𝟐 ∑𝑷 𝒊 𝟏 ∑𝒌 𝟏(𝒃(𝒊,𝒌)) 𝒊 𝟏 𝒌 𝟏
2 alfabet-A
TABLE 2: SHOWS THE NCC VALUE FOR D IFFERENT B AND OF PROPOSED IMAGE WATERMARKING TECHNIQUES USING PROPOSED METHOD (LENA ( AS COVER) & DELL LOGO ( AS WATERMARK) Visibility Factor Α 0.04 0.03 0.025 0.02 0.01 0.005
(5)
Here, PSNR considered for good efficiency is close to 35 dB and the maximum value of NCC is 1 to avoid having a visible watermark but at the same time including the watermark with a large energy to be resistant to attacks.
122
Watermarked Image PSNR 2-Level DWT 1-Level DWT 33.18 33.17 35.43 34.41 37.5052 37.48 39.16 39.15 44.40 44.38 52.05 52.02
Recovered Watermarked Image CC 2-Level DWT 1-Level DWT 99.42 0.9909 0.9936 0.9891 0.9894 0.9828 0.9842 0.9764 0.9581 0.9358 0.8581 0.7873
TABLE 3: SHOWS THE PSNR VALUE FOR D IFFERENT LEVEL OF DWT BASED IMAGE WATERMARKING T ECHNIQUES (BUTTERFLY (AS C OVER) & DELL LOGO ( AS WATERMARK) Visibility Factor Α 0.04 0.03 0.025 0.02 0.01 0.005
Watermarked Image PSNR 2-Level DWT 1-Level DWT 33.18 33.17 35.43 35.41 37.51 37.49 39.17 39.16 44.16 44.38 52.05 52.02
Watermarked Image & Recovered Watermark Image at 0.025 2-Level DWT
TABLE 4: SHOWS THE NCC VALUE FOR D IFFERENT B AND OF PROPOSED IMAGE WATERMARKING TECHNIQUES USING PROPOSED METHOD (BUTTERFLY (AS C OVER) & DELL LOGO (AS WATERMARK) Visibility Factor Α 0.04 0.03 0.025 0.02 0.01 0.005
Recovered Watermarked image CC 2-Level DWT 1-Level DWT 0.9921 0.9862 0.9907 0.9814 0.9848 0.9707 0.9785 0.9586 0.9400 0.8847 0.7541 0.6259
Watermarked Image & Recovered Watermark Image at 0.025 2-level DWT Fig. 6: Shows the Best Result of Watermarked Image and Recovered Watermark Image at 0.025
The comparison of proposed 2-level DWT to 1-level DWT is shown in simulated outcomes. Image Lena is taken as original image and the dell logo image is taken as the watermark image for both the techniques. The size of both the images is 512X512.
The simulated experimental results also evaluated with the visual representation of watermarked and extracted watermark image for human vision system (HVS). Results are clearly seen that the proposed methodology is having the robust efficiency of watermarking with data hiding ability. All experiments are shown the proposed algorithm is efficient in data hiding properties as per HVS.
PSNR Value for different level of DWT based image watermarking techniques (Lena (as cover) &dell logo (as watermark)
V. CONCLUSION
60
In this paper, implementation of watermarking technique has been done using 2-level DWT. In this technique, variable visibility factor is used for the insertion of watermark into the low frequency component of the host image. Simulation results show that the feature of the watermarked image and the recovered watermark are dependent only on the visibility factors and also show that the 2-level DWT give superior results than 1-level DWT.
50 40 30 2-level DWT
20
1-level DWT
10 0
REFERENCES
0.04 0.03 0.025 0.02 0.01 0.005 [1]
VISIBILITY FACTOR Fig. 5: Visual Representation of PSNR Value for Different Levels of DWT at Different Value of Visibility Factor
[2]
The visibility factor (k) is various from 0.05 to 0.009 and we see that as we decrease the value of visibility factor, the value of PSNR increases but at the same time the value of CC decreases hence to get the best result we set the value of visibility factor at 0.025 for both levels as shown in Fig. 6.
[3]
123
P. H. W. Wong, O. C. Au and G. Y. M. Yeung, “A Novel Blind Multiple Watermarking Technique for Images,” accepted by IEEE Transactions on Circuits and Systems for Video Technology: Special Issue on Authentication, Copyright Protection and Information Hiding, Sept. 2003. P V Nagarjuna and K. Ranjeet, 'Robust Blind Digital Image Watermarking Scheme Based on Stationary Wavelet Transform", IEEE IC3, 8-10 Aug. 2013. Samira Lagzian, Mohsen Soryani, Mahmood Fathy, ‘A New Robust Watermarking Scheme Based on RDWT-SVD’International Journal of Intelligent Information Processing, Vol. 2, Number 1, 2011, Doi:10.4156/ijiip.vol2. issue1.3
[4] [5]
[6]
[7]
[8]
[9]
Ezz El-Din Hemdan et al. “Hybrid Digital Image Watermarking Technique for Data Hiding” IEEE 30th national radio science conference 2013, pp220-227, April 2013 Komal Tomar, “A Review Paper of Different Techniques on Digital Image Watermarking Scheme For Robustness” International Journal of Advanced Research in Computer Science and Software Engineering,volume 5,Issue 2, pp 900-904,February 2015 Dr. J. Abdul Jaleel, Jisha Mary Thomas : “Guarding Images Using A Symmetric Key Cryptographic Technique: Blowfish Algorithm“, ISSN: 2277-3754 ISO 9001:2008 Certified International Journal of Engineering and Innovative Technology (IJEIT) Volume 3, Issue 2, August 2013 Mr. Gaurav N Mehta, Mr. Yash Kshirsagar, Mr. Amish Tankariya,“ Digital Image Watermarking: A Review”, International Journal of Scientific Engineering and Technology (ISSN : 2277-1581) www.ijset.com, Volume No.1, Issue No.2 pg:169-174 01 April 2012 Ansu Anna Ponnachen, Lidiya Xavier, “Dwt Based Watermarking For Lifting Based Compression And Symmetric Encryption Of Jpeg Images”, International Journal of Engineering Research & Technology (IJERT) ISSN: 2278-0181 www.ijert.org Vol. 2 Issue 6, June - 2013
[10] [11] [12] [13] [14]
124
Ghazali Bin Sulong, Harith Hasan(Corresponding author), Ali Selamat, Mohammed Ibrahim and Saparudin, “A New Color Image Watermarking Technique Using Hybrid Domain”, IJCSI International Journal of Computer Science Issues, Vol. 9, Issue 6, No 1, November 2012 L. Hu, F. Wan, ‘Analysis on wavelet coefficient for image watermarking’, Int. Conf. MINES’10, pp. 630-634, 2010 Singh A. P., Mishra A., Wavelet Based Watermarking on Digital Image, Indian Journal of computer Science and Engineering 2011. Akhil Pratap Singh,” Wavelet Based Watermarking On Digital Image “Indian Journal of Computer Science and Engineering Vol 1 No 2, 86-91: X. G. Xia, C. G. Boncelet and G. R. Arce, “A Multiresolution Watermark for Digital Images,” in Proc. of IEEE Int. Conf. Image Processing, vol. 1, pp. 548-511, Oct. 1997. E. T. Lin and E. J. Delp, “A Review of Data Hiding in Digital Images,” in Proc. of the Image Processing, Image Quality, Image Capture Systems Conf. (PICS’ 99), pp. 274-278, Apr. 1999.
