C11. Hybrid Digital Image Watermarking Technique for Data Hiding

30th NATIONAL RADIO SCIENCE CONFERENCE (NRSC 2013) April 16-18, 2013, National Telecommunication Institute, Egypt

ell. Hybrid Digital Image Watermarking Technique for Data Hiding

Ezz El-Din Hemdan1, Nawal El-Fishaw/, Gamal Attiya1, Fathi Abd El-Samii 1 2

Computer Science and Engineering Dept., Faculty of Electronic Engineering, Menoufia University, Menouf, Egypt. Electronics and Electrical Comm. Eng. Dept., Faculty of Electronic Engineering, Menoufia University, Menouf, Egypt. E-mails: [email protected]@[email protected] and [email protected]

ABSTRACT: This paper presents a hybrid image watermarking technique for data hiding over Internet. The idea of the proposed technique is based on fusing multiple watermark images using wavelet fusion algorithm. Then, the resultant fused watermark is embedded in the original image using hybrid DWT-SVD watermarking algorithm to produce the watermarked image. The performance of the proposed algorithm is evaluated and a comparative study is done between the hybrid DWT-SVD and SVD watermarking algorithm for single and multiple watermarks. The experimental results verify and prove that the wavelet fusion is an efficient algorithm for fusing multiple watermarks. The image watermarking technique using the hybrid DWT-SVD is more robust than that using the SVD only. The results also prove that the proposed watermarking technique improves both the capacity of the embedded information and robustness without affecting the perceptual quality of the original image. Indeed, the extraction of the fused watermark is possible in the presence of severe attacks. Keywords: Data Hiding, Digital Watermarking, Wavelet Fusion, Discrete Wavelet Transform (DWT), Singular

Value Decomposition (SVD).

I. INTRODUCTION: Recently, the Internet becomes the most important media for information and data communication such as image, audio and video. However, some safety tools should be used to protect the transmission of critical data over the Internet. There has been growing interest in developing techniques to protect data transmission such as encryption and data hiding. An importance sub-discipline of data hiding is called digital watermarking. Digital watermarking is the process of embedding a watermark in a multimedia object. This object may be image, audio, video and any digital content for the purpose of information hiding [1]. Many applications such as E-commerce, E-voting, copyright protection, content authentication, medical safety, broadcasting monitoring, military and indexing can be protected by digital watermarking [2]. Several algorithms have been proposed for watermarking, especially for image watermarking [3]-[6]. In [3], the authors proposed a robust blind digital image watermarking method based on singular value decomposition in wavelet domain to proof of ownership. In [4], a multi-watermarking scheme is proposed by embedding three independent binary watermarks in a grayscale digital image. In [5], a block based digital image watermarking algorithm is developed based on Singular Value Decomposition (SVD) mathematical technique. In [6], a multiple watermarking technique for e-commerce is introduced based on Discrete Wavelet Transform (DWT). Although several algorithms are developed for digital image watermarking to achieve data hiding, a problem arises is that as the capacity of embedded information increases, the quality of the watermarked image and the robustness against attacks are decreased. Therefore, many challenges to design a robust watermarking algorithm are posed due to conflicting of watermarking algorithm characteristics as imperceptibility, robustness, capacity and security [7]. Hence, there is a need for developing a robust image watermarking algorithm with ability to embed multiple watermarks without affecting the perceptual quality of the original image. In this paper, a hybrid digital image watermarking technique is developed for data hiding taking into account increasing capacity without affecting the quality of the watermarked image. In the proposed technique, multiple watermarks images are first fused using wavelet fusion. Then, the resultant fused watermark is embedded in the original image using the hybrid DWT-SVD watermarking algorithm to produce the watermarked image.

978-1-4673-6222-1/13/$31.00 mOl3 IEEE

220


The rest of the paper is organized as follows. Section II provides basic design principles of an image watermarking algorithm. Section III introduces the digital watermarking life cycle while the proposed technique is described in Section TV. Section V describes the proposed algorithm while Section VI presents the experimental results of applying the proposed technique on some images while the paper conclusion is presented in Section VIT.

II. DESIGNING PRINCIPLES OF IMAGE WATERMARKING ALGORITHM Image watermarking algorithm should have the following characteristics [2]: I. TmperceptibiJity: the watermarked image should be not affect the quality of the original image, thus it should be invisible for human eye. 2. Robustness: the watermarked image should not be removed or eliminated by unauthorized person, thus it should resist modifications by attacks. 3. Capacity: the number of bits that can be embedded in the host signal. 4. Security: the watermark should only be detected by authorized person. All these characteristics are often contradictory with each other and must make trade-off between them to produce a robust watermarking algorithm.

III. DIGITAL WATERMARKING LIFE CYCLE The life cycle of digital watermarking is shown in Figure 1. The watermarking technique consists of three steps [8]: 1. Embedding: an algorithm accepts the host and the data to be embedded, and produces a watermarked signal. Then, the watermarked digital signal is transmitted to another person or stored. 2. Attack: unauthorized person try to make modifications. 3. Extraction: is an algorithm which is applied to the attacked signal to attempt to extract the watermark from it. If the signal was unmodified during transmission, then the watermark still is present and it may be extracted. Secure Part Detecting Retrieval Function DR ,

Result�

,

-�...-..-- -.... Fig. I: Digital watennarkin g l ife-cy cl e [8].

IV. PROPOSED TECHNIQUE The proposed hybrid image watermarking technique consists of two phases as shown in Figure 2. In the first phase, the primary watermark is fused with the secondary watermark using wavelet fusion to produce fused watermark. In the second phase, the fused watermark is embedded in the original image using the hybrid DWT SVD watermarking algorithm. On the other hand, the extraction process occurs in two steps; the extraction of fused image and the extraction of both the primary and the secondary watermarks . ............................................................................................................···························_··_··_·_·

Primary Watennark Secondary Watennark

...............................................................

Phase 1

················.....................................................................................................

Original Image Fused Watennark

......................_...._..._..._...._....

r

_---,- --.,------' -

Fig. 2: Proposed Technique.

221

Watennarked Image Phase 2

,


Phase 1: Wavelet Based Image Fusion Wavelet based fusion is an effective technique for combining perceptually important image features which used for many applications such as medical applications and remote sensing applications [9]-[11]. Two images are first transformed using wavelet transform. Then, the fusion is done and finally the inverse wavelet transform IS computed to construct the fused image. This process is shown in Figure 3. Tmage I

---+[

DWT

• Fusion

Tmage 2

---+[

DWT

f

H

IDWT

}---+

Fused Tmage

Fig. 3: Wavel et Fusion of Two Images.

Phase 2: Embedding using Hybrid DWT-SVD Watermarking Algorithm

1) DWT Watermarking Algorithm The Discrete Wavelet Transform (DWT) is currently used to in a wide variety of signal processing applications such as image, audio and video compression and removal of noise in audio. Here one-dimensional DWT separates an image into four bands of data denoted by Lower Level (LL) resolution approximation image, Horizontal Level (HL), Vertical (LH) and Diagonal (HH) detail components [12], As shown below in Figure 4.

LL

HL

LH

HH

Fig. 4: One-l evel DWT.

2) SVD Watermarking Algorithm T he Singular V alue Decomposition (SVD) is mathematical technique. T he main properties of SVD from viewpoint of image processing are the singular V alues (SV s) of image which: 1. Have a good stability when small perturbation is added to an image, its SVs do not change significantly which robust against different attacks. 2. Represent intrinsic algebra properties of image. Using these properties, the watermark can be embedded to SVD matrix without large variation In the watermarked image which robust against many types attacks [13]-[16].

3) Why Hybrid DWT-SVD Watermarking Algorithm? Spatial domain has many disadvantages that do not allow for the exploitation of this subsequent processing in order to increase the robustness of the watermark. It is generally preferable to hide watermarking information in noisy regions and edges of image, rather than in smoother regions. The benefits is two-fold; (1) degradation in smooth regions of an image is more noticeable to Human Visual Technique (HVS), and (2) becomes a prime target for lossy compression. Thus, working in transform domain becomes more attractive [7]. Therefore many techniques are used for image watermarking such as, Discrete Wavelet Transform (DCT), Discrete Wavelet Transform (DWT), Singular Value Decomposition (SVD) and Least Significant bit (LSB).

222


v. ALGORITHM DESCRIPTION

A. Watermarking Embedding Algorithm The watermarking embedding algorithm is shown in Figure 5. The algorithm works as follows: 1. The original image is transformed into four sub-bands using the one-level DWT. 2. The SVD is performed on the A matrix. (1) A = US VT 3. The primary and secondary watermarks are fused using wavelet fusion. 4. The W matrix (fused image) is added to the SVs of the A matrix. (2) D =S + a W To keep the image undistorted, small value of a is chosen (0.01) 5. The SVD is performed on the new modified (D matrix). (3) D = Uw SIr VT" 6. The Aw matrix (watermarked image) is obtained using matrix Sw. (4) A,,=U SIr VT 7. The watermarked image is inversed using the one-levellDWT.

[ ]-"1

Primary Watermark

[

Secondary \Vatermark

+ Fusion

Original Image

Watermarked Image

u,v

Fig. 5: Mul tipl e Watermarking Embedding Strategy .

B. Watermarking Extracting Algorithm The watermarking extracting algorithm is shown in Figure 6. The algorithm works as follows: 1. The watermarked image is transformed using the one-level DWT. * 2. The SVD is performed on the A w matrix (watermarked image). * * *T * (5) A "=U S wV 3. The matrix includes the fused image is computed. * * l' (6) D = u'v S lV V 11' 4. The possibly fused watermark is obtained. * * (7) W = (D - S)/ a 5. Anti-fusion of the fused watermark using wavelet fusion to extract primary and secondary watermarks. 6. The correlation coefficient between the fused watermark and the extracted one is estimated. 7. The correlation coefficient between the primary, secondary watermarks and the extracted ones is estimated.

[

Watermarked Image

H

DWT

H

SVD

1

�S\