In this paper, we propose a novel digital watermarking scheme in DCT domain based fuzzy inference system and the human visual system to adapt the embedding strength of different blocks. Firstly, the original image is divided into some 8×8 blocks, and
of 13
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Similar Documents
  Signal & Image Processing : An International Journal(SIPIJ) Vol.1, No.2, December 2010   DOI : 10.5121/sipij.2010.1210 112 M  AXIMIZING S TRENGTH OF D IGITAL W   ATERMARKS U SING F UZZY L OGIC   Sameh Oueslati 1  and Adnane Cherif  1  and Bassel Solaiman 2 1 Department of Physics: Laboratory of Signal Processing, University of Sciences of Tunis 1060, Tunisia 2 Department of Image and Information Processing, Higher National School of Telecommunication of Bretagne Technopôle, Brest-Iroise, CS 83818, 29238 Brest Cedex 3 France.  A  BSTRACT     In this paper, we propose a novel digital watermarking scheme in DCT domain based fuzzy inference system and the human visual system to adapt the embedding strength of different blocks. Firstly, the srcinal image is divided into some 8×8 blocks, and then fuzzy inference system according to different textural features and luminance of each block decide adaptively different embedding strengths. The watermark detection adopts correlation technology. Experimental results show that the proposed scheme has good imperceptibility and high robustness to common image processing operators.  K   EYWORDS    Digital Watermark, Fuzzy logic, Co-occurrence Matrix, Robustness &Medical imaging. 1.   I NTRODUCTION   The necessity of fast and secure diagnosis is vital in the medical world. Nowadays, the transmission of visual data is a daily routine and it is necessary to find an efficient way to transmit them over networks [13], [15], [17]. The main objective is to guarantee the protection of medical images during transmission, and also once this digital data is archived [6], [7]. The subsequent challenge is to ensure that such coding withstands severe treatment such as compression [26]. When a physician receives a visit from a patient, he often requires a specialist opinion before giving a diagnosis. One possible solution is to send images of the patient, along with a specialist report, over a computer network [33], [1]. Nevertheless, computer networks are complex and espionage is a potential risk. We are therefore faced with a real security problem when sending data. Forethical reasons, medical imagery cannot be sent when such a risk is present, and has to be better protected [11]. Watermarking is the best form of protection in cases such as this. Many different techniques for the watermarking of text already exist. Recently, watermarking has been proposed for medical information protection. Even though most of the work on watermarking has concerned medical images in order to verify image integrity or improve confidentiality [32], watermarking also provides a new way to share data. Basically, watermarking is defined as the invisible embedding or insertion of a message in a host document, an image, for example. Anand et al . [9], proposed to insert an encrypted version of the electronic patient record (EPR) in the LSB (Least Significant Bit) of the gray scale levels of a medical image. Although the degradation in the image quality is minimum, the limitations and fragility of LSB watermarking schemes is well-known. Miaou et al . [23] proposed a method to authenticate the srcin of the transmission, the message embedded is an ECG, the diagnosis report and physician’s information. Macq and Dewey [24] insert information in the headers of medical images. These approaches are not robust against attacks such as filtering, compression, additive noise, etc. neither to geometrical attacks such as rotation or scaling transformations. In this work, we propose the use of DICOM metadata as a watermark to embed in medical images  Signal & Image Processing : An International Journal(SIPIJ) Vol.1, No.2, December 2010 113 extracted from the DICOM file. We introduce a metric of an image quality evaluation ( wPSNR ). This distortion metric, that has no relation with the content characteristics of the image fits to the HVS and therefore is more suitable for digital Watermarking. In this method The FIS and the HVS combined are used to adjust the watermarking strength. In order for the embedded watermark to be even more robust against different types of attacks, it is essential to add as powerful invisible watermark as possible. Based on the above reasons, the proposed scheme in this paper divides the srcinal image into some 8×8 blocks, Feature Extraction by Co-occurrence Matrix, and the FIS according to different textural features and luminance of each block decide adaptively different embedding strengths. Finally, the watermark extraction is semi-blind, i.e., it is accomplished without using the srcinal image which makes the approach higher security. As a result, the watermark is more robust and imperceptible. The remainder of the paper is organized as follows. Section 2 provides a detailed a description of the HVS model and FIS, Section 3 describes the method for embedding and extraction watermark. In Section 4, the experimental results and comparisons are shown. The conclusions of our study are stated in Section 5. 2.   H UMAN V ISUAL S YSTEM AND T EXTURAL F EATURES S ELECTION   2.1. Human Visual System (HVS) in Watermarking It is known that there is a trade off between the imperceptibility and robustness of a digital watermarking system. Up to now, the watermark's invisibility issue is only tackled by the embedding depth [22], [10]. But, this action is very limited because the watermark signal, once mapped in the media space, is spread all over the image. Uniform areas of the image are very sensitive to watermark addition so they only support extremely small embedding depth, whereas edge areas, for instance, support deeper watermark addition. The new spatial masking is built according to the image features such as the brightness, edges, and region activities.  With the same watermark embedding, the quality of watermarked image using the proposed adaptive masking is much better than the one without using the adaptive masking. The human visual system has been characterized with several phenomena that permit pixel element adjustments to elude perception. 2.2. Feature Extraction by Co-occurrence Matrix A general procedure for extracting textural properties of image was presented by Haralick et al . [2]. Each textural feature was computed from a set of COM  probability distribution matrices for a given image. The COM measures the probability that a pixel of a particular grey level occurs at a specified direction and a distance from its neighbouring pixels. Formally, the elements of a GG ×   gray level co-occurrence matrix d  P   for a displacement vector ),( dydxd   = is defined as: (1) Where (.,.)  I  denote an image of size  N  N   ×   with G   gray values ),( sr  ,  N  N vt   ×∈ ),( , ),(),( dysdxr vt   ++= and . is the cardinality of a set. ),(  ji p  refers to the normalized entry of the co-occurrence matrices. That  R jiP ji p d   / ),(),(  = , where  R   is the total number of pixel pairs ),(  ji . For a displacement vector ),( dydxd   = and image of size  M  N   ×  R   is given by ))(( dy M dx N   −− . The eight nearest-neighbour resolution cells (3 by 3 matrix), which define the surrounding image pixels, were expressed in terms of their spatial orientation to the central pixel ),(  ji called a reference cell [3]. The eight neighbours represent all the image pixels at a distance of 1. For example, resolution cells ),1(  ji + and ),1(  ji  − are }{  jvt  I isr  I vt sr  jiP d   === ),(,),(:)),(),,((),(  Signal & Image Processing : An International Journal(SIPIJ) Vol.1, No.2, December 2010 114 the nearest neighbours to the central cell ),(  ji  in the horizontal direction )0( 0 = θ   and at a distance )1(  = d  . This concept is extended to the three additional directions )135,90,45( 0 = θ   as well as when a distance equals 2, 3 and so on. Of the 14 features, we found that the following 5 have the most powerful discrimination ability as texture features of the composite images: angular second moment (  ASM  ), contrast ( CON  ), correlation ( COR ), variance (VAR) and entropy (  ENT  ). Using the formulas of the textural features, the angular second moment, contrast, correlation, variance and entropy are presented in Table 1. « Angular second moment » is a measure of homogeneity of an image. The higher value of this feature indicates that the intensity varies less in an image. « Contrast » measures local variation in an image. A high contrast value indicates a high degree of local variation. « Correlation » is a measure of linear dependency of intensity values in an image. For an image with large areas of similar intensities, correlation is much higher than for an image with noisier, uncorrelated intensities. « Variance » indicates the variation of image intensity values. For an image with identical intensity for all images, the variance would be zero and «entropy» is an indication of the complexity within an image. A complex image produces a high entropy value. Table 1. A set of statistical features. 2.3. Mamdani Fuzzy Inference System (FIS) Fuzzy Inference Systems (FIS) are popular computing frameworks based on the concepts of fuzzy set theory, which have been applied with success in many fields [8]. Their success is mainly due to their closeness to human perception and reasoning, as well as their intuitive handling and simplicity, which are important factors for acceptance and usability of the systems [30]. In figure1 the general schema of a FIS is portrayed. In particular, three main modules are of particular interest: a fuzzifier, a rule base and a defuzzifier. While the fuzzifier and the defuzzifier have the role of converting external information in fuzzy quantities and vice versa, the core of a FIS is its knowledge base, which is expressed in terms of fuzzy rules and allows for approximate reasoning [28]. Typically, a FIS can be classified according to three types of •   The angular second moment (  ASM  ) { } 2 ),( ∑∑ = ij  ji p ASM   (2) •   The contrast ( CON  ) ∑∑ = ij  ji p jiCON  ),(),( 2  (3) •   The correlation ( COR )  y x ji y x S S mm jiijp COR ..),( , ∑  −=  (4) •   The entropy (  ENT  ) ∑∑ −= ij  ji p ji p ENT  ),(log),(  (5) •   The variance (VAR) ∑∑  −= ij  ji p ji piVAR ),()),(( 2 (6)  Signal & Image Processing : An International Journal(SIPIJ) Vol.1, No.2, December 2010 115 models that are distinguished in the formalization of the fuzzy rules [29]. In this work, we focus on the Mamdani type, which is characterized by the following fuzzy rule schema: Figure 1. Scheme of a Fuzzy Inference System The main feature of such type of FIS is that both the antecedents and the consequents of the rules are expressed as linguistic constraints [17]. As a consequence, a Mamdani FIS can provide a highly intuitive knowledge base that is easy to understand and maintain, though its rule formalization requires a time consuming defuzzification procedure. As aforementioned, the proposed adaptive watermarking scheme computes a watermark weighting function using a HVS and a FIS. In order to efficiently extract the masking information, while taking into account the local characteristics of the image. In what follows we present the insertion algorithm whose steps are detailed. In this work, the FIS is utilized to compute the optimum watermark weighting function that would enable the embedding of the maximum-energy and imperceptible watermark. This FIS is therefore ideal to model the watermark weighting function, as it can incorporate the fuzzy and nonlinear aspect of human vision. In this work, the input membership function was divided into three linguistic values, where each input denoted as low average and high respectively. The determination of the membership function is done by using the help of ANFIS Toolbox in MATLAB. This technique enabled excellent model development for non-linear process in which the rules were automatically generated under ANFIS environment. 3.   W ATERMARK E MBEDDING AND D ETECTION   Here, we describe a frequency domain based watermarking system with blind detection. Basically, it comprises a watermark embedding and a detection process. Figure 2 illustrates the watermark embedding process. A host image is first transformed using one of the available transformation tools; our work in the DCT is used. Coefficients are then selected for watermark insertion. After embedding process, an inverse transformation is applied. To obtain a watermarked image. The blind detection process is shown in Figure 3. It is performed by means of a correlation function. A   possibly corrupted image is transformed using the same tool as at the embedding process. Among the proposed activities in the domain of medical image watermarking, the algorithm [31], which is encoded on a pair of frequency values {0, 1}. Use of frequency domain DCT can fulfill not only the invisibility through the study of optimizing the insertion gain used, but also security by providing a blind algorithm or use the srcinal image No is not essential and the extraction of the mark is through a secret key [5], [12]. Fuzzifier Rule base Defuzzifier Ouput Input  Signal & Image Processing : An International Journal(SIPIJ) Vol.1, No.2, December 2010 116 Figure 2. The watermark embedding process In fact, the human eye is more sensitive to noise in lower frequency components than in higher frequency ones. However, the energy of most natural images is concentrated in the lower frequency range, and watermark data in the higher frequency components might be discarded after quantization operation of lossy compression. In order to invisibly embed the watermark that can survive lossy data compressions, a reasonable trade-off is to embed the watermark into the middle-frequency range of the image. In our approach we have two key insertion to secure the place where the watermark was introduced. The first key tells us the positions of the two coefficients selected with the same value of quantization [18], [20]. While the second key position Relates blocks which bear the marks among all the components total blocks transformed image. Phase extraction is as follows: Compare the values of DCT coefficients to determine if the respective bit of the message was a "0" or a "1". Block DCT decomposition Choice of two coefficients of each block Inverse of block DCT decomposition Watermarked image Read the mark and turn it into vector )( i M    0)(  = i M  Applied α  − to the first coefficient and α  + to the second coefficient Applied α  + to the first coefficient and α  − to the second coefficient Host image
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!