A New approach of SVD-DWT Video Watermarking Embedding Algorithm

Transcription

1 A New approach of SVD-DWT Video Watermarking Embedding Algorithm Pawandeep Kaur, Sonika Jindal M.Tech Student, Assistant Professor, Department of CSE SBS State Technical Campus Ferozepur, India Abstract Digital watermarking is used to protect digital content such as images, audio and videos that have been tampered maliciously and with higher accuracy. In this paper, to maintain the quality of video and to ensure the ownership we propose a new SVD- DWT watermarking embedding technique. Singular value decomposition (SVD) is an important transform technique in robust digital watermarking.we apply the 3 level DWT and SVD on selected frames and embed the watermark into randomly selected frames with the help of secret key to authenticate the video by considering the video quality, robustness and video imperceptibility. Keywords Digital video watermarking, Performance evaluation metrics, Secret key, DWT SVD algorithm. I. INTRODUCTION A rapid growth in multimedia (audios, videos, images) and illegal transfer of this multimedia content over the internet are becoming important issues in digital era. Digital watermarking is used to protect this sensitive information using different watermarking technologies. Video watermarking is relatively a new technique in multimedia technology. [1] Video watermarking is the process in which watermark is embedded in a video sequence by using a secret key and then extraction is performed at the other end using the same secret key as shown in Fig: 1. The embedded watermark should be robust against variety of attacks, so that video can be protected from illegal copying and provide security against several attacks such as frame dropping, frame swapping and frame averaging [2].The two types of watermark can be used such as visible watermark and invisible watermark. We can add the watermarks either in the whole frames of video or in certain frames depending upon the requirement [3]. Video watermarking is mainly used in two domains: spatial domain, frequency domain. [4] Depending upon the various Fig. 2. Applications Of Video Watermarking applications, video watermarking is used in fingerprinting, copyright protection,video authentication, copy control and broadcast monitoring as shown in Fig: 2. The rest of the paper is organized as section 2 defines the Related Works. Section 3 describes the Performance Evaluation Metrics. Section 4 describes Proposed Architecture. Section 5 describes Proposed Algorithm.Section 6 defines Experimental Results. II. RELATED WORK The techniques involved in video watermarking have been grouped into three major categories; spatial domain watermarking techniques, frequency domain watermarking techniques and format based watermarking techniques. The first category is spatial domain watermarking in which watermark is embedded in frames by directly modifying the pixel values of that frame or replacing the bits of selected frame pixels [5]. In second category [5] Frequency domain watermarking techniques, first coefficients of transformed video frames are modified and then transformations are applied and at last the inverse process is applied to get the watermarked video. In third category format based video watermarking, a number of MPEG-2 and MPEG- 4-based techniques on GOP modification, high frequency DCT coefficient manipulation and DCT block Fig. 1. A General Video Watermarking Process 36

2 classification are proposed. Compression in block-based schemes like MPEG-2 is achieved by using forward and bi-directional motion prediction. Just another watermarking system was initially developed by Philips for broadcast monitoring and DVD applications. [6] In embedding process, secret key is used to generate the distributed reference pattern (Pr) and then reference watermark (Wr) is created according to following equation. Wr = Pr shift (Pr, message) (1) On the other side, an extractor will recognize the watermark precisely identify the source clip and also the time, location and channel of the broadcast. Spread spectrum technique is a very effective method for embedding watermark for compressed video [7]. [8] Describes that watermark bits are repeated N/4 times, where N denotes number of pixels to be watermarked and then spread information bits are then modulated with a crypto logically secure pseudo noise signal, scaled and added to the image or video pixels. [9] designed a technique in which watermark is directly embedded and detected in VLC domain for copyright protection.video decoding stage, a video bit stream is usually decoded into codeword by VLC decoding.[10] proposed a frame by frame video watermarking technique in which first S S watermark is created and then pattern is repeatedly embedded. [11] Presented a linear collusion analysis of watermarked digital video along with analytic notion of statistically invisible video watermark. [12] Proposed a method in order to increase robustness against signal processing and lossy compression in which watermark is embedded into perceptually significant components. CDMA has a more natural application in the watermarking of uncompressed digital video. Digital video is modeled as a bit plane stream along the time axis. Using a modified m-sequence, bit planes of specific order are pseudo randomly marked for watermarking. Then, the required watermark is mapped to a single bit plane and spread via another 2D m-sequence, along a stream parallel to that of the video [13]. Perceptual watermarks are designed to exploit properties of the human visual system in order to provide extremely high-quality original content. [14] In order to achieve transparency and robustness to variety of attacks there will be requirement of more sophisticated use of perceptual information in the watermarking process. Watermarking algorithm based on group of frames (GOF) helps to maintain temporal imperceptibility requirements. [15] proposed a method that generates the message authentication code ( MAC) for a group of frames (GOF ) using coefficients from the last but one high pass band at full level of temporal wavelet decomposition. Some techniques based on digital signature uses temporal wavelet transform for the generation of message authentication code. A. DWT Watermarking: divides an image into two sections such as in lower resolutions as well as in higher resolutions. Lower resolution means LL components and higher resolution means horizontal (HL), vertical (LH) and diagonal (HH) detail components. The low frequency part is further divided into two sections of high and low frequencies. This process is repeated number of times to compute multiple scale wavelet decomposition [16]. [17] Proposed a method in which different parts of watermark are embedded into different video scenes using DWT transform and this method provide robustness against variety of attacks such as frame dropping, frame averaging and statistical analysis. [18] Proposed a method in which 3D DWT is applied using perceptual mask and embedding is performed by weighing the mark through the defined mask and then the Inverse 3D DWT (IDWT) is performed. 37 Advantages: More accurate model because its properties similar to HVS and more robust to noise addition. Disadvantages: Sometimes, higher frequencies distort the quality of image. B. SVD Watermarking : is a mathematical tool which decomposes a matrix into two orthogonal matrices and one diagonal matrix consisting of the singular values of the matrix [19].The SVD mathematical technique provides an elegant way for extracting algebraic features from an image and improves watermark robustness and resistance against many kinds of attacks. SVD is an effective method to split the system into a set of linearly independent components. A digital Image X of size MxN can be represented by its SVD as follows: X = USV T U U 1, U 22..Um V V 1, V 22. Vn S = σ 1 0 σ 2 (2) Where U is an MxM orthogonal matrix, V is an NxN orthogonal matrix, and S is an MxN matrix in which only the diagonal elements represents the singular values. T denotes the transpose of the matrix. III. PERFORMANCE EVALUATION METRICS The performance of the watermarked images must be evaluated by using some quality measures such as MSE, SNR, PSNR and BER. 1) The MSE (mean square error): [20] defined it as average squared difference between a reference image and a distorted image. It is calculated as: MSE = 1 XY X Y i=1 j=1 (c i, j e(i, j)) 2 X represents the height and Y represents the width of the image. c(i,j) and e(i,j) are the respective pixel value of the original image and embedded image.

3 2) SNR (Signal to Noise ratio):it is defined as the ratio of signal power to the noise power. It is expressed in decibels. A ratio higher than 1:1 (greater than 0 db) indicates more signal than noise. It measures the signal strength relative to the background noise (Unwanted signal). It is calculated by the formula given below: SNR db =10log 10 P signal N noise 3) The PSNR (peak signal to noise ratio):it is a quality metric used to determine the degradation in the embedded image with respect to the host image or also defined as ratio between maximum power of a signal and power of distorted signal [20]. It is most easily defined via the mean squared error (MSE) as: PSNR = 10log 10 L L MSE L denotes the peak signal value of the cover image which is equal to 255 for 8 bit images. 4) The BER (bit error ratio): [20] Defined it as the ratio that describes how many bits received in error over the number of the total bits received. It is often expressed as percentage and calculated by comparing bit values of embedded image and cover image. BER = P/(H W) H and W are the respective height and width of the water-marked image. P is the count number whose initial value is zero and it increments by one if there is any bit difference between cover and embedded image. 5) The SSIM (structural similaraity index ): Structural similarity index is a method for measuring the similarity between two images. SSIM is designed to improve on traditional methods like peak signal to noise ratio (PSNR) SSIM(x,y)= (2μ x μ y +c1)(2σ xy +c2) μ x 2 +μ y 2 +c1 (σ x 2 +σ y 2 +c2) With μ x the average of x, μ y the average of y and σ x 2 the variance of x, σ y 2 the variance of y. IV. PROPOSED ARCHITECTURE The proposed method effectively hides the secret data into video using existing video watermarking techniques. Fig: 3 give a complete overview of data flow in proposed algorithm. This method uses some frames of video to hide the secret data. The frames selected to hide secret data are random frames and not sequential frames. Hence each frame that contains the secret data can be identified using secret key, a 10 digit number provided by user. The careful selection of frames is done by using several functions that are made up from secret key. So, watermark is embedded in whole video and not in some parts of video. V. PROPOSED ALGORITHM In this section, we have discussed some motivating factors in design of our approach to watermarking. We have used DWT and SVD for developing the algorithm. Among various tools, SVD and DWT are more reliable in digital watermarking. Due to the fact of localization in both spatial and frequency domain, wavelet transform is the most preferable transform among all other transforms. After converting the video into frames, we have applied 3 levels DWT on selected frames. In Fig. 3. Proposed Architecture and mean squared error (MSE), which have proven to be inconsistent with human eye perception. It is calculated by formula given below: Fig level DWT the next stage, the SVD is applied to selected sub -bands as shown in Fig: 4, and embed the same original watermark by modifying the singular values. Embedded watermark in middle frequencies increases the robustness to variety of attacks. The procedure of embedding a digital watermark into the original video is depicted in Fig. 3. At last, inverse SVD and inverse DWT is applied in order to reconstruct the watermarked digital video. A. Watermark embedding algorithm Apply DWT to the selected frames repeatedly up to the third level. 38

4 Perform SVD transform on approximation and all the detail parts in third level of wavelet transform, f Q= U Q S Q V T Where Q {LL3, LH3, HL3, HH3}. f Q = U Q S Q V T (1) Perform SVD transform on watermark, W = U W S W V T W (2) In general, embedded watermark at this stage. Modify the singular values of approximation and all the detail parts with the singular values of the watermark as: γ Q = γ Q + α Q γ W (3) Here, is scale factor of combined transform, which value is 0.05 in this paper. Take inverse combined transform and reconstruct the watermarked video. VI. EXPERIMENTAL RESULTS The main focus of this algorithm is its dynamic and key dependent frame selection technique. We have implemented and experimented it using MATLAB.The experimental results are as below which show original frames and corresponding watermarked frame. We test the proposed watermarking algorithm with different variations using colored host video clips. Each video clip is partitioned into different number of frames. We employed rhinos video sequence in AVI format and frame rate fps. Each video clip consists of 114 frames. We have selected 10 random frames and embed watermark such as penguins.jpg of size ( ) in that frames. The 10 random original frames and their corresponding watermarked frames are shown in Fig: 4. Watermarked Video quality was estimated by SSIM, PSNR, BER and MSE. Video Rhinos Fig. 7. Watermarked Frames TABLE I. (Experimental results) Frame no. SSIM PSNR BER MSE We then tested the robustness and quality of watermarked video using different performance evaluations metrics. For each frame we have calculated the SSIM, PSNR, MSE, BER and MSE as shown in above table. After calculating the SSIM, PSNR, BER and MSE for each frames, we calculate the average value of all 4 metrics for all selected frames which is shown in following Fig: 8. Fig. 5. Original Frames Fig. 6. Watermark Image Fig. 8. Average value of SSIM,PSNR,BER and MSE The experimental results and average value calculated shows that propose DWT-SVD based video watermarking algorithm is robust. The calculated PSNR value is 52.93db which shows quality of watermarked video appear visually identical to the original one and there is no degradation in visual quality. 39

5 VII. CONCLUSION This paper provides that proposed algorithm is dynamic and key dependent which gives the robustness and good quality watermarked video. In future this algorithm can be expanded by watermark extraction process using same secret key which has been used during watermark embedding and we can perform this with/without attacks and by using dynamic key verification and authentication technique. ACKNOWLEDGMENT This is to express my sincere gratitude to Dr. Satvir Singh, Associate Professor, Department of Electronic & Communication Engineering, SBS State Technical Campus, Ferozepur (Punjab), India, for sparking in me the enthusiasm and initiative to discover and learn. I am truly thankful to him for guiding me through the entire paper and being as a motivator in this learning curve. REFERENCES [1] Jayamalar, T and Radha, V, Survey on digital video watermarking techniques and attacks on watermarks, International Journal of Engineering Science and Technology, vol. 2, Pp , [2] Singh, Amit and Jain, Susheel and Jain, Anurag, A Survay: Digital Video Watermarking. [3] Madia, Jigar and Dave, Kapil and Sampat, Vivek and Toprani, Parag, Video Watermarking using Dynamic Frame Selection. [4] Do{\"e}rr, Gwena{\"e}l and Dugelay, Jean-Luc, A guide tour of video watermarking, Signal processing: Image communication, Elsevier, vol. 18, Pp ,2003. [5] Jayamalar, T and Radha, V, Survey on digital video watermarking techniques and attacks on watermarks, International Journal of Engineering Science and Technology. J. Magn. Japan, vol. 2, Pp , [6] Kalker, Ton and Depovere, Geert and Haitsma, Jaap and Maes, Maurice J, Video watermarking system for broadcast monitoring, International Society for Optics and Photonics, Pp , [7] Shoemaker, Chris, Hidden bits: A survey of techniques for digital Hartung, Frank H and Girod, Bernd, Digital watermarking of raw and compressed video, International Society for Optics and Photonics, Pp ,1996. [8] Hartung, Frank H and Su, Jonathan K and Girod, Bernd, Spread spectrum watermarking: Malicious attacks and counterattacks, International Society for Optics and Photonics, Pp ,1999. [9] Lu, Chun-Shien and Chen, Jan-Ru and Liao, H- YM and Fan, Kuo-Chih, Real-time MPEG2 video watermarking in the VLC domain, IEEE, vol.2, Pp , [10] Su, Karen and Kundur, Deepa and Hatzinakos, Dimitrios, Novel approach to collusion-resistant video watermarking, International Society for Optics and Photonics, Pp , [11] Su, Karen and Kundur, Deepa and Hatzinakos, Dimitrios, Statistical invisibility for collusionresistant digital video watermarking, Multimedia, IEEE Transactions on, vol.7, Pp , [12] I. J. Cox, J. Kilian. F. T. Leighton and T. Shamoon, Secure spread spectrum watermarking for multimedia, IEEE transactions on image processing, vol. 6, Pp , [13] Mobasseri, Bijan G, Exploring CDMA for watermarking of digital video, International society for optics and photonics, Pp , [14] Wolfgang, Raymond B and Podilchuk, Christine I and Delp, Edward J, Perceptual watermarks for digital images and video, Proceedings of the IEEE, vol.87, Pp , [15] Upadhyay, Saurabh and Singh, Sanjay Kumar, Video Authentication-An Overview, International Journal of Computer Science and Engineering Survey, vol 2, [16] Potdar, Vidyasagar M and Han, Song and Chang, Elizabeth, A survey of digital image watermarking techniques, Industrial Informatics, [17] INDIN' rd IEEE International Conference on, Pp , [18] Chan, Pik-Wah and Lyu, Michael R, A DWTbased digital video watermarking scheme with error correcting code, Springer, Pp , [19] Campisi, Patrizio and Neri, Alessandro, Video watermarking in the 3D-DWT domain using perceptual masking, IEEE, vol.1, Pp.I--997, [20] K.-L. Chung, W.-N. Yang, Y.-H. Huang, S.-T. Wu, Y.-C. Hsu, On svd-based watermarking algorithm, Applied Mathematics and Computation Pp 54 57, [21] A. K. Singh, N. Sharma, M. Dave, A. Mohan, A novel technique for digital image watermarking in spatial domain, in: Parallel Distributed and Grid Computing (PDGC), nd IEEE International Conference on, IEEE, 2012, pp