Digital Image Watermarking using Hybrid DWT-DCT-SVD Approach based on JND

 

Manjusha Tikariha1, Amar Kumar Dey2

1Research Scholar, BIT, Durg (C.G.)

2Sr. Assistant professor, BIT, Durg (C.G.)

*Corresponding Author Email: manjusha.etc@gmail.com, amardeybit@gmail.com

 

ABSTRACT:

The digital revolution in digital image processing has made it feasible to create, manipulate and transmit digital images in an effortless and fast manner. The adverse affect of this is that the similar image processing techniques can be used by hackers to tamper with any image and use it illegitimately. This has made digital image safety and integrity the top prioritized issue in today’s information explosion. Most popular technique that is used for copyright protection and authentication is watermarking. Digital watermarking is an unpretentious and effective approach to afford copyright fortification. Watermark transparency is obligatory primarily for copyright protection. Based on the characteristics of human visual structure, the just noticeable distortion (JND) can be used to substantiate the transparency requirement. DCT based watermarking offers compression while DWT based watermarking techniques offer scalability. SVD based algorithm offer a robust method of watermarking with minimum or no distortion. The DWT and SVD have some advantages for the digital watermark, which can be good to resist geometric attacks and frequency domain attacks. Thus all these desirable properties can be utilized to create a new robust watermarking technique. This paper aims at developing a hybrid image watermarking algorithm which satisfies both imperceptibility and robustness requirements. In order to achieve objectives this research used singular values of Transformation’s (DWT-DCT) sub bands to embed watermark. Further to increase and control the strength of the watermark, this work used an efficient JND model. An optimal watermark embedding method is developed to achieve minimum watermarking distortion. Experimental results are provided in terms of Peak signal to noise ratio (PSNR), Mean Squared Error (MSE) and Correlation to demonstrate the effectiveness of the proposed algorithm. This paper shows an analysis of all the above method.

 

KEYWORDS: Just Noticeable Distortion; Digital Watermarking; multi-scale embedding; Wavelet subspaces; Singular Value Decomposition; Discrete Wavelet transform; Discrete Cosine Transform.

 

 

INTRODUCTION:

Digital media offers several distinct advantages over analog media, such as high quality, easy editing, high performance and easy duplication. The high spreading of broadband networks and new developments in digital technology has made ownership protection and authentication of digital multimedia a very important issue.

 

Digital watermarking provides a possible solution to the problem of easy editing and duplication of images, since it makes possible to identify the author of an image by embedding secret information in it.

 

Digital watermarking technique is one of the solutions to avoid unauthorized copying or tampering of multimedia data. Recently many watermarking schemes have been proposed to address this problem. Digital Watermarking is defined as the process of hiding a piece of digital data in the cover data which is to be protected and extracted later for ownership verification [1], a basic watermarking process is shown in figure 1. Some of the important applications of watermarking technique are copyright protection, ownership verification, finger printing, and broadcast monitoring. The features of watermarking include robustness and perceptibility. Robustness indicates the resistivity of watermark against different types of attacks such as cropping, rotating, scaling, low pass filtering, resizing, and addition of noise, JPEG compression, sharpness, histogram equalization and contrast adjustment. Those attacks are either intentional or unintentional. Robustness is the property which is important for ownership verification whereas the fragility is important for image authentication.

 

Figure 1. Flow of Watermarking process

 

Robustness of watermarking algorithm is obtained to a maximum level when information is hidden in robust components of cover data. The increasing perceptibility will also decrease the quality of watermarked image. The watermarking schemes are broadly categories into two main domains i.e. spatial domain and the transform domain. In spatial domain watermarking the watermark is embedded by directly modifying the intensity values of the cover image. The most popular technique is the least significant bit (LSB) method. In transform domain the watermark is embedded by modifying the frequency coefficients of the transformed image.

 

The common methods in the transform domain are Fourier transform (DFT), discrete cosine transform (DCT), discrete wavelet transform (DWT), etc. Recently, singular value decomposition (SVD) was explored for watermarking. It is one of the most useful numerical analysis techniques having property that the singular values (SVs) of an image do not change significantly when a small perturbation is added to an image. [2-5].

 

Just-noticeable distortion (JND), which refers to the maximum distortion that the human visual system (HVS) cannot perceive, plays an important role in perceptual image and video processing. In comparison with JND estimation for images, estimation of the JND profile for video needs to take into account the temporal HVS properties in addition to the spatial properties. In most circumstances, the human visual system (HVS) makes final evaluations on the quality of images and video that are processed, transmitted, and displayed.  Thus, it is essentially futile to spend significant effort on encoding those signals that are beyond the human perception. Just noticeable distortion (JND), which accounts for the maximum distortion that the HVS does not perceive, can serve as a perceptual threshold to guide an image/video processing task. In image compression schemes, JND can be used to optimize the quantizer or to facilitate the rate-distortion control. Information of higher perceptual significance is given more bits and preferentially encoded, so that the resultant image is more appealing.

 

It is well known that there are three main mutually conflicting properties of information hiding schemes: capacity, robustness and indefectibility [6]. It can be expected that there is no a single watermarking method or algorithm with the best quality in the sense that three mentioned above properties have the maximum value at once. But at the same time it is obvious that one can reach quite acceptable quality by means of combining various watermarking algorithms and by means of manipulations in the best way operations both in the spatial and in the frequency domains of an image. In paper [7] an approach to combining of DWT and DCT to improve the performance of the watermarking algorithms, which are based solely on the DWT, is proposed. Watermarking was done by embedding the watermark in the first and second level DWT sub-bands of the host image, followed by the application of DCT on the selected DWT sub bands. The combination of these two transforms improved the watermarking performance considerably when compared to the DWT-only watermarking approach. As a result this approach is at the same time resistant against copy attack. In addition, the fragile information is inserted in a way which preserves robustness and reliability of the robust part.

 

Robustness is the property which is important for ownership verification whereas the fragility is important for image authentication. Robustness of watermarking algorithm is obtained to a maximum level when information is hidden in robust components of cover data. The increasing perceptibility will also decrease the quality of watermarked image. Generally information could be hidden, directly by modifying the intensity value or pixel value of an image or its frequency components [8]. The former technique is called spatial domain technique and later is called frequency domain technique. To obtain frequency components of an image, it needs to be transformed using any one of the transformation techniques such as Discrete Fourier Transformation (DFT), Discrete short Fourier transformation (DSFT), Discrete Cosine Transformation (DCT) [9][10], Walsh Hadamard transformation (DHT) [11][12], and Discrete wavelet Transformation (DWT)[13][14][15][16]. In Transform domain casting of watermark can be done in full frequency band of an image or in specific frequency band such as in low frequency band or in high frequency band or in middle frequency band.

 

Digital Image Watermarking

The growth of the Internet along with the increasing availability of multimedia applications has spawned a number of copyright issues. One of the areas that this growth has fueled is that of digital watermarking. Digital watermarking is the general technique of embedding a blob of information in the original file, such that an altered file is obtained. The blob of information, thus included, serves one of different uses, such as, identifying piracy, sensing tampering, or reassuring integrity. The approaches to watermarking are diverse and can be broadly classified based on their visibility, robustness, or fragility. Their uses are also versatile, as they can be applied to text, images, audio, or video.

 

Watermarking techniques can broadly be classified based on their inherent characteristics: visible and invisible.

 

Visible watermarks:

A visible alteration of the digital image by appending a “stamp” on the image is called a visible watermark. This technique directly maps to that of the pre-digital era where a watermark was imprinted on the document of choice to impose authenticity.

 

Invisible watermarks:

By contrast, an invisible watermark, as the name suggests that this is invisible for the most part and is used with a different motive. While the obviousness of visible watermarking makes distinguishing legitimate and illegitimate versions easy, its conspicuousness makes it less suitable for all applications. Invisible watermarking revolves around such suitable factors that include recognizing authentic recipients, identifying the true source and non-repudiation.

 

Another way of classifying watermarking technique is a factor of its usage: robust, fragile, or semi-fragile, and spatial or spectral watermarks.

 

Robust watermarks:

Watermarks can be used to hold knowledge of ownership. Such watermarks need to remain steadfast to the original image to do what they advertise. The intactness of the watermark is a measure of its robustness. These watermarks must be able to withstand normal manipulations to the image such as reduction of image size, lossy compression of image, changing the contrast of the images, etc.

 

Fragile watermarks:

These are complementary to robust watermarks and are, as a rule, more change-sensitive than robust watermarks. They lose their mettle when they are subject even to the smallest changes.

 

Figure 2. Classification of watermarking techniques

 

Their use lies in being able to pin-point the exact region that has been changed in the original watermarked image. The methods of fragile watermarking range from checksums and pseudo-random sequences in the LSB locale to hash functions to sniff any changes to the watermark.

 

Semi-fragile watermarks:

These watermarks are a middle ground between fragile watermarks and fragile watermarks. They engulf the best of both worlds and are more resilient than fragile ones in terms of their robustness. They also are better than robust watermarks in terms of locating the regions that have been modified by an unintended recipient.

 

Requirements for watermarking algorithms

A watermarking algorithm should be consistent over following properties and parameters:

·         Transparency: The most fundamental requirement for any Watermarking method shall be such that it is transparent to the end user. The watermarked content should be consumable at the intended user device without giving annoyance to the user.

·         Security: Watermark information shall only be accessible to the authorized parties. Only authorized parties shall be able to alter the Watermark content.

·         Ease of embedding and retrieval: Ideally, Watermarking on digital media should be possible to be performed “on the fly”. The computation need for the selected algorithm should be minimum.

·         Robustness: Watermarking must be robust enough to withstand all kinds for signal processing operations, “attacks” or unauthorized access.

·         Effect on bandwidth: Watermarking should be done in such a way that it doesn’t increase the bandwidth required for transmission.

·         Interoperability: Digitally watermarked content shall still be interoperable so that it can be seamlessly accessed through heterogeneous networks and can be played on various plays out devices that may be watermark aware or unaware.

 

Overview of Singular Value Decomposition

Singular value decomposition is a linear algebra technique used to solve many mathematical problems [29]. The theoretical background of SVD technique in image processing applications to be noticed is [30]:

 

a)      The SVs (Singular Values) of an image has very good stability, which means that when a small value is added to an image, this does not affect the quality with great variation.

b)      SVD is able to efficiently represent the intrinsic algebraic properties of an image, where singular values correspond to the brightness of the image and singular vectors reflect geometry characteristics of the image.

c)      An image matrix has many small singular values compared with the first singular value. Even ignoring these small singular values in the reconstruction of the image does not affect the quality of the reconstructed image.

 

Just-Noticeable Distortion (JND):

The model that is given below is used for the calculation of spatial domain Just Noticeable Distortion (JND) profile for a given image. This model can calculate the JND profile for a 4 level DWT decomposition.  Denotes the waveletcoefficient at position  of decomposition level  and orientation Note thatr=0 stands for first level of decomposition.

 

The texture is calculated from overall four bands of DWT and the value of texture is the sum of values of texture in three bands (LH, HL, and HH) and value of variance in LL band.

 

The value of JND from above information is calculated as:

 

Proposed Watermarking Scheme:

This paper develops a hybrid digital image watermarking algorithm which satisfies both imperceptibility and robustness requirements. In order to achieve objectives this research used singular values of Transformation’s (DWT-DCT) sub bands to embed watermark. Further to increase and control the strength of the watermark, this work used an efficient JND model. First both the cover image and watermark are transformed through discrete wavelet transform, then the sub-bands is further transformed through discrete cosine transform. Then the proposed model calculated the singular values of transformed part of both cover image and watermark image using singular value decomposition. After that, the singular values are combined with each other by taking JND factor between cover image and watermark image in order to evolve a more promising approach in field of digital image watermarking.

 

The watermarking schemes proposed here are combined DCT/DWT based processes, where the benefits of DWT are taken into consideration in choosing the most proper sub-band for watermark embedding in order to provide both robustness and imperceptibility and hence the sub-band is chosen after performing one level DWT on the host image. Secondly, DCT is applied on the DWT sub-bands and for watermark embedding purpose the middle frequency DCT coefficients are selected to provide further robustness to the schemes.

 

The Wavelet Series is just a sampled version of continuous WT and its computation may consume significant amount of time and resources, depending on the resolution required. The Discrete Wavelet Transform (DWT), which is based on sub-band coding is found to yield a fast computation of Wavelet Transform. It is easy to implement and reduces the computation time and resources required. The DWT is computed by successive low-pass and high-pass filtering of the discrete time-domain signal. This is called the Mallat algorithm or Mallat-tree decomposition. Its significance is in the manner it connects the continuous-time multi-resolution to discrete-time filters.

 

The DWT decomposes input image into four components namely LL, HL, LH and HH where the first letter corresponds to applying either a low pass frequency operation or high pass frequency operation to the rows, and the second letter refers to the filter applied to the columns. At each decomposition level, the half band filters produce signals spanning only half the frequency band. This doubles the frequency resolution as the uncertainty in frequency is reduced by half. The 2D-DCT can not only concentrate the main information of original image into the smallest low frequency coefficient, but also it can cause the image blocking effect being the smallest, which can realize the good compromise between the information centralizing and the computing complication. So it obtains the wide spreading application in the compression coding.

 

Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD)

DWT decomposes images into four bands. In this work, the watermark is embedded in maximum energy bands (means all frequencies). And after that, all bands are further transformed using discrete cosine transform, these results in robustness to a wide range of attacks. SVD is an optimal matrix decomposition technique. It packs maximum energy into as few coefficients as possible. SVD has the ability of adapting to variations in local statistics of an image. So, watermarking schemes using SVD are typically of large capacity. Different types of wavelets are also used in experiment including ‘db1’ or Haar Wavelet, ‘db2’ or Daubechies level two wavelets, ‘db4’ or Daubechies level four wavelets, ‘sym2’ or Symlets level two wavelets, ‘dmey' or Discrete Meyer wavelets and ‘db45’ or Daubechies level 45 wavelets

 

A human visual model for DCT-DWT-SVD domain image Watermarking

A relationship between a hybrid DCT-DWT domain human visual model and the modification threshold of singular values is established. The threshold, which in image adaptive, is used to determine the watermarking strength and guarantees the imperceptibility of the watermark.

 

Watermark Embedding

Let ‘A’ be the cover image. Apply DWT to decompose the image into four sub-bands LL, HL, LH and HH. Take any of these four sub-bands. Apply DCT to the chosen sub-band. Let ‘B’ denote the matrix obtained after applying DCT. Now B acts as the host image. Apply SVD so that ‘B’ can then be written as B = UB ΣB VBT where UB and VB T are the orthonormal unitary matrices of B. The term ΣB constitutes the singular values of the matrix of B.

 

Let ‘W’ represent the watermark. Apply DWT and take any of the four sub-bands. Apply DCT to the chosen sub-band. Let ‘S’ denote the matrix obtained after applying DCT. Now B acts as the host image. Apply SVD so that ‘S’ can then be written as S = USΣSVST where US and VST are the orthonormal unitary matrices of S . The term ΣS constitute the singular values of the matrix S.

 

After calculating singular values of both the cover image and watermark . We have to combine them using perceptual factor of JND as:

 

Here,  is the JND value calculated as per DWT. Then singular value S is again combined with the unitary matrix of cover image in order to restore it. Finally, inverse DCT and inverse DWT is performed to produce the watermarked image.

 

Figure 3: Block Diagram for Watermark Embedding Algorithm

 

 

Watermark Extraction

Let ‘A’ be the cover image. Apply DWT and take any of the four sub-bands. Apply DCT to the chosen sub-band. Let ‘B’ denote the matrix obtained after applying DCT. Now B acts as the host image. Apply SVD so that ‘B’ can then be written as B = UB ΣB VBT where UB and VBT are the orthonormal unitary matrices of B. Term ΣB constitutes the singular values of the matrix of B.

 

Let ‘w*’ be the watermarked image. Apply DWT and take any of the four sub-bands. Apply DCT to the chosen sub-band. Let ‘A*’ denote the matrix obtained after applying DCT. Now A* acts as the host image. Apply SVD so that ‘B’ can then be written as B = UA*ΣA*VA*T where UA* and VA*T are the orthonormal unitary matrices of A*. Term ΣA* constitutes the singular values of the matrix of A*. Finally, the watermark is extracted from the selected wavelet coefficients by applying the singular value decomposition on the watermarked image and restoring it with the help of same JND coefficients generated earlier. Watermark is extracted by subtracting the singular values obtained above.

 

Figure 4: Block Diagram for Watermark Extraction Algorithm.

 

PERFORMANCE EVALUATION

In this section the results of our study is shown. Several experiments are done to evaluate the effectiveness of the presented watermarking algorithm. In our experiments 512×512 color Barbara image was taken as the cover image and color logo was used as the watermark. Figure 3.(a) and 3.(b)., show original image and watermark image which are used in this experiment, respectively. The Figure 4.(a) and 4.(b)., depict watermarked image and watermark which extracted after no attack, respectively.

                    (a)                                        (b)

Figure.3. (a) Host image; (b) Watermark used in the experiment.

 

                  (a)                                          (b)

Figure.4. (a) Watermarked image; (b) Watermark which extracted after no affect.

 

Imperceptibility test:

Imperceptibility property must be preserved by watermarking scheme. In order to compare the cover image and watermarked image peak signal to noise ratio (PSNR) is used. PSNR in decibels (dB) is given below:

 

Where E is Mean Square Error, is pixel value of original image  of watermarked image and its logarithmic unit is dB given by Formula:

If PSNR value is greater than 35dB the watermarked image is within acceptable degradation levels, i.e. the watermarked is almost invisible to human visual system. Average PSNR obtained by our proposed method is 51dB.The MSE and PSNR values obtained for two set of images i.e., Lena-logo and Barbara-logo which are tabulated in Table I.

 

Table I: PSNR and MSE for the case of invisibility (by comparing input cover image and watermarked image)

Cover Image

Watermark Image

MSE

PSNR

Barabara

Logo

0.468

51.4267dB

Lena

Logo

0.369

52.5209dB

 

Robustness test:

The robustness against various attacks is test. The attacks are performed on the watermarked image and watermark is extracted from this watermarked image using the extraction algorithm described in section IV. Correlation is used to measure the similarity between the original (W) and extracted watermarks (W*). The correlation factor NHS may take values between 0 (random relationship) to 1 (perfect linear relationship). In general, a correlation coefficient of about 0.7 or above is considered acceptable.

In order to get the measure of the robustness of the presented algorithm, several image processing attacks are implemented on the watermarked image.The watermarked image has been tampered with the built-in functions of MATLAB software suite. The attacks performed in this research work are as follows: Image Adjustment, Histogram Equalization, Rotation, Salt and pepper noise attack, Gaussian noise attack, Speckle noise. The performance analysis results are cited in Table II.

 

Table II: PSNR and MSE for the case of robustness (by comparing input and output watermark

Attacks

The barbara test image

Correlation

Proposed Method

No Attack

1

Gaussian Noise

2.5%

.9289

5 %

.87791

10 %

.81589

Salt and pepper noise addition

2%

.98188

5%

.95332

10%

.91175

Speckle Noise

2.5%

.98064

5%

.96113

10%

.92715

Image rotation

10 clockwise

.99994

 

400 anticlockwise

1

Sharpened

.98587

 

CONCLUSION:

In this paper, a hybrid technique which is a combination of DCT-DWT-SVD is implemented using MATLAB. In this method, the watermark is embedded very deep into the cover image since three transform (DCT, DWT, SVD) are taken before embedding the watermark which help in resilience the attacks. This method can be used for copyright protection, tamper detection, fingerprinting, authentication and secure communication. The proposed scheme is robust to, noise adding attacks, sharpness adjustment attack, rotation attack and other signal proceeding attacks. Better robustness is obtained. The MSE and PSNR values obtained are found to be good. Experimental results are presented to claim the robustness and correctness of the proposed watermarking process.

 

REFERENCES:

1      F. Hartung and M. Kutter, "Multimedia Watermarking Techniques," in Proc. of the IEEE, vol. 87, no. 7, pp. 1079-1107, July 1999.

2      R. Sun, H. Sun, and T. Yao, “A SVD and quantization based semifragile watermarking technique for image authentication,” in Proc.Int. Conf. Signal Process. pp. 1592–1595, (2002)

3      C. C. Chang, P. Y. Tsai, and M. H. Lin, “SVD-based digital image watermarking scheme,” Pattern Recogn. Lett. 26, 1577–1586, (2005).

4      J. M. Shieh, D. C. Lou, and M. C. Chang, “A semi-blind digital watermarking scheme based on singular value decomposition,” Comput. Stand. Inter. 28, 428–440, (2006).

5      R. Liu and T. Tan, “An SVD-based watermarking scheme for protecting rightful ownership,” IEEE Trans. Multimedia 4, 121–128, (2002).

6      Chih-Chin Lai and Cheng-Chih Tsai.(November 2010) ‘Digital Image Watermarking Using Discrete Wavelet Transform and Singular Value Decomposition’, IEEE Transactions On Instrumentation And Measurement Vol.59, No.11, pp. 3060-3063.

7      Chun-Shien Lu and Hong-Yuan Mark Liao.(October 2001) ‘Multipurpose Watermarking for Image Authentication and Protection’ IEEE Transactions On Image Processing Vol. 10, No.10, pp.1579- 1592.

8      V. Potdar, et al "A Survey of Digital Image Watermarking Techniques", in Proc.of the EEE International Conference on Industrial Informatics, Australia Perth, pp. 709-716, 2005.

9      Liu Quan, AI Qingsong “A combination of DCT based and SVD based watermarking, ICSP proceedings of IEEE International conference on signal processing, pp. 873-876, 2004.

10    Feng Liu, et al. “A Watermarking Algorithm for Digital Image Based on DCT and SVD” Congress on Image and Signal Processing, 2008.

11    Emad E. Abdallah, et al. “A robust block-based image watermarking scheme using fast Hadamard transform and singular value decomposition” proceedings of The 18th International Conference on Pattern Recognition, pp:673 - 676 , 2006.

12    Tang Xianghong, Yang Lianjie, Yue Hengli, Yin Zhongke, “A Watermarking Algorithm Based on the SVD and Hadamard Transform”, Proceedings of International Conference on Communications, Circuits and Systems,Volume 2, pp. 874-877 , 27-30 May 2005.

13    Ali Al-Haj, “A Hybrid Digital Image Watermarking Algorithm”, 4th International Conference on Innovations in Information Technology. pp: 690 - 694 Nov. 2007.

14    Liu Liang and Sun Qi “A new SVD-DWT composite watermarking”, ICSP proceedings of IEEE International conference on signal processing .2006

15    Jung-Chun Liu, Chu-Hsing Lin, and Li-Ching Kuo” A Robust full band image watermarking scheme” Proceedings on IEEE. 2006

16    Qiang Li, et al, “Adaptive DWT-SVD Domain Image Watermarking Using Human Visual Model” proceedings of 9th international conference on advanced communication Technology, Volume 3, pp: 1947 - 1951, Feb.2007.

17    Ahmed Bouridane ,Lahouari Ghouti ,Mohammad K. Ibrahim and Said Boussakta.(April 2006 ) ‘Digital Image Watermarking Using Balanced Multiwavelets’ IEEE Transactions On Signal Processing Vol. 54, No. 4, pp. 1519-1536.

18    Chih-Chin Lai and Cheng-Chih Tsai.(November 2010) ‘Digital Image Watermarking Using Discrete Wavelet Transform and Singular Value Decomposition’, IEEE Transactions On Instrumentation And Measurement Vol.59, No.11, pp. 3060-3063.

19    H. Andrews and C. Patterson, "Singular Value decompositions and Digital Image Processing," IEEE Trans. on Acoustics, Speech,and Signal Processing, vol. 24, no. 1, pp. 26-53, Feb. 1976.

20    Ke-feng he, et al, “Watermarking for images using the HVS and SVD in the wavelet domain” Proceedings of IEEE International Conference on Mechatronics and Automation, pp.2352- 2356, 2006

21    N. Jayant, “Signal compression: technology targets and research directions,” IEEE Journal on Selected Areas in Communications, vol. 10, pp. 314–323, 1992.

 

Received on 28.03.2015                                  Accepted on 30.04.2015        

©A&V Publications all right reserved

Research J. Engineering and Tech. 6(2): April-June, 2015 page 293-300

DOI: 10.5958/2321-581X.2015.00045.8