[IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) -...
Transcript of [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) -...
![Page 1: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/1.jpg)
Digital Watermarking Method based on Fuzzy Image Segmentation Technique
Methaq Gaata 1,2, William Puech 1, Sattar Sadkhn2 I LIRMM UMR CNRS 5506, University Montpellier 2, 161, rue Ada 34392 Montpellier, France
2 Computer Science Department, Babylon University, Hilla Babylon 00964, Iraq
Abstract- In this paper, we propose new watermarking method for image copyright protection in spatial domain based on fuzzy image segmentation technique. The fuzzy image approach is employed to determine appropriate locations to hiding watermark by identifYing the features map (edges regions) which are less sensitive to the human visual system. Our method optimizes robustness and imperceptibility which are known to be inversely proportional to each other. Transparency performance of proposed method was evaluated by four image quality measures are PSNR, SSIM and two similarity measures which are based on the fuzzy logic Sj, S2' The evaluation results demonstrate the high robustness of the proposed algorithm against the common image processing operators. Our method is also is efficient as it only uses simple operations such image fuzzijication and fuzzy image segmentation technique.
Keywords- digital watermarking, fuzzy logic, fuzzy image processing.
I. INTRODUCTION
Information hiding became an important field as the use of
the Internet became popular. Information hiding is a young field and it is growing in an exponential rate [1, 2]. In digital images the information hiding applications could be divided into two groups is steganography and digital watermarking. As a branch of information hiding, digital watermarking provides an effective approach for copyright protection, authentication, security and integrity of digital media data.
The basic procedure for digital image watermarking is to embed some hidden information into image data, while the quality of the watermarked data is retained, and the watermark can still be detected under different kinds of intentional and unintentional attacks, i.e., watermarking should be robust, but the ways of pursuing imperceptibility and robustness are conflict. It is an important issue to find a fair balance between imperceptibility and robustness. The robustness is the ability to resist certain malicious attacks, such as the general operations of image processing (clipping, filtering, enhancement, compression, etc.) which is important issue in digital watermarking. The Imperceptibility is original image should not be visibly degraded by the watermark. In
other words, we must ensure that an unauthorized user do not perceive the existence of the watermark. Imperceptibility
ensures the excellent perceptual quality of the protected image [3].
Watermarking techniques can be divided into various categories in numerous ways. In the case of still digital images, there are three primary methods for insertion and extraction of a watermark. These are spatial domain, transform domain and color space methods. In spatial domain
978-1-4673-0753-6/11/$26.00 ©2011 IEEE
technique [4], the watermark embedding is achieved by directly modifying the pixel values of the host image. The most commonly used method in the spatial domain technique
is the least significant bit (LSB). In transform domain
technique [5], the host image is first converted into frequency
domain by transformation method such as the discrete cosine transform (DCT), discrete Fourier transform (DFT) or discreet wavelet transform (DWT) ,etc. then, transform domain coefficients are modified by the watermark. The inverse transform is [mally applied in order to obtain the watermarked image. In color space watermarking method, the image is converted from RGB color space to different color space such as the YUV color space [6].
In recent years, a number of adaptive watermarking methods for digital image have been proposed. Especially, the adaptive watermarking technique for incorporating the
features of the human visual system (HVS) model can provide an excellent solution [7, 8]. Recently, intelligent algorithms, such as neural networks and fuzzy methods are introduced into digital watermarking technique and can
simultaneously improve robustness and visual quality of the watermarked image. In [9], Lou et al. proposed an adaptive watermarking method using fuzzy logic technique. Wu et al. [10] proposed a digital watermarking scheme in DCT domain
based on fuzzy clustering technique, which can adaptively control the embedding strength of different blocks. Sakr et al. [11] proposed an adaptive image watermarking algorithm
based on dynamic fuzzy inference system. In [12], Chang et al. proposed a Fuzzy-ART based adaptive digital watermarking scheme in DCT domain. Lunde et al. [13]
proposed novel hardware for an adaptive encrypted watermarking method based on fuzzy logic. In [14], Oueslati et al. proposed an adaptive watermarking algorithm
performed in the wavelet domain which exploits a human visual system (HVS) and a Fuzzy Inference System (FIS).
In this paper, we propose a new watermarking algorithm based on fuzzy logic for copyright protection. The fuzzy
image segmentation approach is employed to determine appropriate locations to hiding watermark. Based on the fact
that human eye is less sensitive to changes in textured areas than in smooth areas. This algorithm embeds the watermark in the feature pixels (edges components) of the cover image. Thus, it requires less number of computations.
This paper is organized in the following manner. The concept of fuzzy image processing is presented in Section 2. The watermark inserting and detecting method is proposed in Section 3. Experimental results are shown in Section 4. Section 5 concludes this paper.
218
![Page 2: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/2.jpg)
II. FUZZY IMAGE PROCESSING
Fuzzy image processing is not a unique theory. It is a collection of different fuzzy approaches to image processing. Nevertheless, the following definition can be regarded as an attempt to determine the boundaries:
Fuzzy image processing is the collection of all approaches that understand, represent and process the images, their segments and features as fuzzy sets. The representation and processing depend on the selected fuzzy technique and on the
problem to be solved [15].
The most important of the needs of fuzzy image processing are as follows:
1. Fuzzy techniques are powerful tools for knowledge representation and processing.
1. Fuzzy techniques can manage the vagueness and ambiguity efficiently.
2. In many image-processing applications, we have to use expert knowledge to overcome the difficulties (e.g., object recognition, scene analysis).
Fuzzy image processing has three main stages: image fuzzification, modification of membership values, and, if necessary, image defuzzification. The general of fuzzy image processing is shown in the Fig. 1.
Fig. 1. The general structure of fuzzy image processing.
The fuzzification and defuzzification steps are due to the fact that we do not possess fuzzy hardware. Therefore, the
coding of image data (fuzzification) and decoding of the results (defuzzification) are steps that make possible to process images with fuzzy techniques. The main power of fuzzy image processing is in the middle step (modification of membership values, see Fig. 2). After the image data are transformed from gray-level plane to the membership plane (fuzzification), appropriate fuzzy techniques modify the membership values. This can be a fuzzy clustering; a fuzzy rule-based approach, a fuzzy integration approach, and so on [16].
978-1-4673-0753-6/11/$26.00 ©2011 IEEE
0; 50 55 63 Original image > ., ..:! c; 58 208 199 :h ...::: '" Q. .. 218 100 89 0
I 0.19 0.21 0.25
Fuzzification .,
0.23 0.8 0.76 ., c; 0.82 O A 0.33 '" Q..
Modification of Q. memberships �
S ...::: 0.09 0.07 0.25 -" =
0.10 0.92 0.81 � D efuzzification 0.94 0.65 0.51
\11 0; 19 0.17 67 Result image 6 0) 25 242 201 �.§ E Q. 247 143 128 0
Fig. 2. Steps of fuzzy image processing.
III. PROPOSED WATERMARKING ALGORITHM
The algorithm we propose here is based on applying the fuzzy image segmentation on the digital image in which a
watermark is to be embedded. In this section, we describe the main steps of the proposed watermark embedding and extraction algorithms. The flowchart of the proposed watermarking algorithm is presented in Fig. 3.
Watermark Bits
Original Image
Noise Removal using Fuzzy Filter
Image Segmentation using Fuzzy Method
Watermarked Image
Fig. 3. Flowchart of proposed watermarking method.
219
![Page 3: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/3.jpg)
A . Image juzzijication
Firstly, fuzzification is used to transform original image (crisp data set) into a fuzzy image (fuzzy data set). For fuzzification, we use a fuzzy gaussian function. Thus in fuzzy set theory, an image X of size M x N having L gray levels of pixels can be considered as an array of fuzzy singletons, each associated with a membership value [17]. Thus a fuzzy image can be represented as:
X = lJl)xij I f.1(xij). (1)
I J
The membership function of the gray value essentially reflects the membership or belongingness of the pixel to a certain class. The equation for the fuzzy Gaussian function is:
(.j -jlx(x-fii IiXj=e . (2)
The inputs to the equation are fi the spread and h the midpoint. The spread generally ranges from 0.01 to 1, with the larger the value results in a steeper distribution around the midpoint. (The default value is 0.1). The default value for midpoint is calculated as the mean of the input data.
B. Noise removal
A fundamental problem of image analysis is to effectively remove noise from an image while keeping intact its fundamental structure constituting of edges and comers. In this step, fuzzy filter is used for the noise removal from image.
Let the gray level of the pixel at the location (m, n) be given by Xmn. If the central pixel gray level be given by Xc. then the membership value f.1 (xu) of any pixel in the
neighborhood of Xc is given by [17]:
[-(X -X. } ] [d2] f.1(xij) = exp
c a 1J exp � . (3)
where d is the distance between the central pixel X and the
neighboring pixel xij' a and f3 are two scaling factors which
determine the extent of flatness of the membership function. The membership values serve as the weights in the averaging process of the pixels in the neighborhood W. Fuzzy filter algorithm can be applied as follows: first, for a central pixel Xc. determine a neighborhood W of some suitable size. Second, for each pixel xij in the neighborhood determine the
membership value Ii (xu). Third, the modified gray value of
the central pixel can thus be given by:
978-1-4673-0753-6/11/$26.00 ©2011 IEEE
(4)
C. Fuzzy image segmentation
In this step, we build features map to find appropriate regions for embedding watermark. We use thresholding approach based on fuzzy edge detection for image segmentation to compute values of features map. The main
idea of features map is to partition the image into two
regions: Edges and non-Edges. Image segmentation process is used to partitioning image
into multiple regions or sets of pixels. Actually, partitions are different objects in image which have the same texture or color. The result of image segmentation is a set of regions that collectively cover the entire image or a set of contours extracted from the image. Fuzzy segmentation divides pixels into fuzzy sets i.e. each pixel may belong partly to many sets and regions of image [15, 17].
The problem of thresholding involves identifying an optimal threshold T and segmenting the scene into two meaningful regions-objects (0) and background (B), such as:
0= {(m,n) lImn � T}
B = {(m,n) lImn < T} (5)
Fuzzy thresholding involves partitioning the image into two
fuzzy sets 0 and jj corresponding to object and background by
identifying the membership distributions 116 and Iljj associated with these regions. Here, we use the fuzzy
thresholding scheme which minimizes the fuzziness in the thresholded description and at the same time accommodates the variations in the gray-values of each region. The membership function used to indicating the degree of edginess in each neighborhood defines as:
1[1+ if;n ;-1 if fmn � T j./cpmn}=
o if1mn<T
(6)
where v is the mean gray-value of 0 and the parameter k controls the amount of fuzziness in the segmented description. A similar membership assignment was employed
for jj also. They classified the pixels unequivocally into object or background regions with the help of the hard threshold T.
220
![Page 4: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/4.jpg)
D. Watermark embedding
Edges play an important role in the perception of images, since the HVS cannot distinguish the changes in edge regions. We therefore embed watermark in regions that are "Edges "according to the features map.
After computing the membership values from the fuzzy segmentation operation as explained in equation (6). We can extract the fuzzy features map (edges). The feature values in interval [0-1] will be used to determine the places in which the watermark bits will be stored in the original image. This
is done by checking membership value of the fuzzy segmentation operation if membership value is greater than mean value of membership values then store the watermark bit in the same position in original image. This operation continues until the end of the watermark bits. Inserting the watermark in these selected pixels is done by changing the specific bit of the pixels. The watermark bits are embedded in pixels of original image which satisfy the following condition:
(7)
where Mm represents mean value for membership values of the fuzzy segmentation operation.
(al)
(a2)
E. Watermark e xtraction
Watermark extraction is the inverse process of watermark embedding. We only need to know the threshold value and the secret key. We use membership value file with the same assumptions as described in equation (6). The steps are image fuzzification, noise removal and image segmentation. The watermark bits are extracting from pixels watermarked image that should satisfy the following condition:
(8)
IV. EXPERIMENTAL RESULTS
In order to evaluate the proposed watermarking algorithm, several images from the IVe 2009 image database [18] are used. Fig. 4.al and Fig.4.a2 represent the original images. Fig. 4.b 1 and Fig. 4.b2 represent the watermarked images of
original images. We evaluate the proposed method in terms of
imperceptibility and robustness. About transparency evaluation, we use two similarity measures are based on the fuzzy logic [19].
(bl)
(b2)
Fig. 4. (al&a2) Original images, (bl&b2) Watermarked image
978-1-4673-0753-6/11/$26.00 ©2011 IEEE 221
![Page 5: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/5.jpg)
1 ( ] M N r )r
SlA,B)=I- - I II,uA(x,y)- ,uB(x,y)1 MN x=]y=]
(9)
where r>O. The similarity measure S2 is based on the fuzzy minkowski distance, and the observation that the smaller the distance between original image A and watermarked image B.
SlA, B) = IAn BI
IAUBI
Iff! min(,u A(x,y)"u B(x,y)) MN Ix,y=l max(,u A (x,y),,u B(x,y))
(10)
where flA(x,y) and flB(x,y) are fuzzy membership values of
the pixel of the two images A and B respectively. The quality evaluation results of proposed watermarking algorithm by using Sj, S], PSNR and SSIM [14] measures are shown in Table 1.
Table 1. Quality evaluation for proposed algorithm.
Quality Image 1 Image 2
measures
SI 0.931 0.905
S2 0.921 0.881
PSNR (dB) 43.991 42.94
SSIM 0.922 0.832
The robustness evaluation of the proposed algorithm is
performed with some attacks from Stirmark and Checkmark benchmarks. We try to apply different kinds of operation to the watermarked data and extract the watermarking information. After extracting watermark, we use similar measurement of the extracted and the referenced watermarks as validation; it can be defined by the Normalized Correlation (NC). The formula which calculates similarity is defined as:
(11)
*
where Wk and Wk are referenced and extracted watermarks
and L is length of watermark bits. Here, the size of watermark is (200 bits) and the watermark repeatedly inserted
into the original image. The robustness evaluation is reported in Table 2.
V. CONCLUSIONS
In this paper, we have proposed a watermarking algorithm based on fuzzy logic which can be used for image copyright.
978-1-4673-0753-6/11/$26.00 ©2011 IEEE
Table 2. Robustness evaluation by NC values.
Attack Types Image 1 Image 2
Jpeg compression 3% 0.918 0.923
Jpeg compression 15% 0.875 0.889
Jpeg compression2000 3% 0.865 0.876
Jpeg compression2000 15% 0.811 0.824
Gaussian noise (J =50% 0.915 0.93
Blurring 0.882 0.895
Winner filtering 0.916 0.931
Median filtering 0.972 0.981
Sharping 0.964 0.978
Experimental results show that the degradation by embedding the watermark is too small to be visualized. Therefore, the proposed watermarking scheme can achieve a better tradeoff between the robustness and the transparency. In addition, the robustness of this algorithm is strong enough to some attack like gaussian noise, compression and blurring. Future work
will focus on making the watermarks robust to more attacks and optimizing the algorithms to provide higher capacity and robustness.
REFERENCES
[I] F. P. Petitcolas, R. 1. Anderson, and M. G. Kuhn, "Information hiding-a survey," Proceedings of the IEEE, special issue on protection of
multimedia content Vol. 87, No.7, pp.1062-1078, 1999. [2] N. Johnson and S. Jajodia, "Exploring steganography: seeing the
unseen," iEEE Computer Vol. 58, No.8, pp.26-34, 1998. [3] S. Mabtoul, E. Tbn-Elhaj and D. Aboutajdine, "A blind chaos- based
complex wavelet domain image watermarking technique," International Journal on Computer Science and Network Security, Vol. 6, No. 3, March 2006.
[4] W. Lu, H. Lu, and F.-L. Chung, "Robust digital image watermarking based on subsampling," Applied Mathematics and Computation, Vol. 181,pp. 886-893,2006.
[5] A. A. Reddy and B. N. Chatterji, "A new wavelet based logowatermarking scheme," Pattern Recognition Letters, vol. 26, pp. 1019-1027,2005.
[6] H. Ren-Junn, K. Chuan-Ho, and C. Rong-Chi, "Watermark in color image," Proceedings of the first International Symposium on Cyber
Worlds, pp. 225-229,2002. [7] W. Lu and L. Hongtao, "Novel robust image watermarking based on
subsampling and DWT," Multimedia Tools and Applications, DOl 10.1007/s1 1042-01 1-0794-1, 2011.
[8] H. Peng, J. Wang, and W. Wang, "Adaptive image watermarking approach based on kernel clustering and HVS," WiLF '09 Proceedings
of the 8th international Workshop on Fuzzy Logic and Applications, LNAT 5571,213-220, 2009.
[9] D.C. Lou, and T.L. Yin, "Adaptive digital watermarking using fuzzy logic techniques," Opt. Eng. 41,2675,2002; [doi:10.1l17/1.1499968].
[10] J. Wu and J. Xie, "Adaptive image watermarking scheme based on HVS and fuzzy clustering theory," iEEE Int. Con! Neural Networks & Signal Processing 2, pp. 1493-1496,2003.
[II] Sakr, Zhao, and Croza, "A dynamic fuzzy logic approach to adaptive HVS-based watermarking," iEEE International Workshop on Haptic Audio Visual Environments and Their Applications, pp.121-126, 2005.
[12] Chang, and Zhang, "Fuzzy-ART based adaptive digital watermarking scheme," iEEE Trans. on Circuits and Systems for Video Technology Vol. 25, No.1, pp.65-81, 2005.
222
![Page 6: [IEEE 2011 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT) - Bilbao, Spain (2011.12.14-2011.12.17)] 2011 IEEE International Symposium on Signal](https://reader030.fdocuments.us/reader030/viewer/2022020300/575094f71a28abbf6bbdb13f/html5/thumbnails/6.jpg)
[13] P. U. Lande, S. N. Talbar, and N. Shinde, "Robust image adaptive watermarking using fuzzy logic an FPGA approach," International Journal of Signal Processing, Image Processing and Pattern
Recognition Vol. 3, No.3, pp.97-107,2010. [14] S. Oueslati, A. Cherif, B.Solaiman, "A fuzzy watermarking system
using the wavelet technique for medical images," International Journal
of Research and Reviews in Computing Engineering I, pp.43-48, 20 II. [15] H. R. Tizhoosh, Fuzzy Image Processing,Springer-Verlag, ISBN: 3-
540-63137-2,2004. [16] S. N. Sivanandam, S. Sumathi and S. N. Deepa, Introduction to Fuzzy
Logic using MATLAB, Springer-Verlag Berlin Heidelberg, New York 2007.
[17] T. Acharya and A. K. Ray, Image Processing Principles and
Applications, Published by John Wiley & Sons, Inc., Hoboken, New Jersey, 2005.
[18] Available online at: hup:llwww.irccyn.ec-nantes. fr/-autrusse/Databases/LARI.
[19] Z. Wang, A. C. Bovik, H. R. Sheikh and E. P. Simoncelli, "Image quality assessment: From error visibility to structural similarity," IEEE
Trans. Image Processing, Vol. 13, No. 4, pp. 600-612, 2004.
978-1-4673-0753-6/11/$26.00 ©2011 IEEE 223