Blind Video Watermarking Using Multilevel
Hybrid Regression Technique under Compression
Based Attacks
Sakshi Batra Chandigarh Group of Colleges, Mohali, India
Email: [email protected]
Rajneesh Talwar Chandigarh Group of Colleges, Jhanjeri, India
Email: [email protected]
Abstract—Whirlwind neoplasm of networks such as
Internet, wireless communication and Intranet has
facilitated the wide use of multimedia techniques and digital
data. Transmission of information has become rather swift
and sheer due to the same but as all good things come at a
price, this in turn has multiplied the casualties of
bushwhack and assail on the data that is being transmitted.
To fence this, many techniques have come into foray in the
recent past. These techniques are broadly categorized under
Copyright Protection Solutions. The main features of
information hiding are capacity, security and robustness.
Video watermarking usually prefers robustness. In a robust
algorithm to eliminate, the watermark without rigorous
degradation of cover content is not possible. This paper
introduce frequently used key techniques and Video
Watermarking with features required to robustly
watermark a video for a valuable application, algorithm is
based on Singular Value Decomposition (SVD) and
redundant wavelet transform. The algorithm showed high
level of imperceptibility, when compared with base
approach; performance varied with respect to robustness
and payload.
Index Terms—digital video watermarking, Singular Value
Decomposition (SVD), robustness, imperceptibility, DWT,
DCT
I. INTRODUCTION
Selling and Marketing of art works is easier using
multimedia techniques and internet but it also comes with
the hard fact that such procedure becomes vulnerable to
anti business activities like copying [1]. Thus Copyright
Protection is of utmost importance to facilitate
widespread and fool proof use of such advanced
technologies. Digital Watermarking is one such
technique that fire walls content owners from
mischievous elements [2]. Various types of information
is embedded in digital content using Digital
Watermarking. In simpler terms, a watermark is used as
information to validate the data and protect copyrights [3].
Manuscript received May 13, 2015; revised January 30, 2016.
Associated with the widespread circulation of videos are
issues of copyright infringement, authentication and
privacy. One possible solution is to embed some invisible
information into the videos where the embedded
information can be extracted for different purposes.
Digital watermarking is a process to embed some
information called watermark into different kinds of
media called Cover Work [4]. While some watermarks
are visible, most watermarks of interest are invisible.
There are many classes of invisible watermarks for
different applications such as fragile watermarks and
robust watermarks. Fragile watermarks are designed to be
broken easily by video processing operations. The broken
watermark serves as an indication of alteration of the
original video. Major applications include tampering
detection of videos placed on the WWW and
authentication of videos received from questionable
sources [5]. Robust watermarks are required to remain in
the watermarked video even after it has been attacked.
Digital Watermarking could also be considered as a
prolongation of Steganography, which has built its base
as an encouraging solution for copyright protection [6]. It
assesses superior control over embedded information.
The purpose of Digital Watermarking is that it does not
leave any conspicuous tag on the content which helps it
sustain its esteem. It is not possible to discard such
watermarking without demoting the content giving this
technique an upper hand over Cryptography. Such digital
watermarks are impalpable and could only be detected by
fit virtuoso.
The attacks may also be casual or unintended attacks
which are common video processing such as filtering,
compression, scaling, cropping, etc. Major applications
include ownership establishment, copyright and
distribution control. Data hiding watermarks, also called
steganography [7], are used to embed data in the videos
with the intention to have the data recovered perfectly at
the receiver. Such methods usually assume that there are
no hostile or even casual attacks. Data integrity is not
secure in image transfers. The copyright data may be in
the form of text [8].
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 569doi: 10.18178/ijeee.4.6.569-574
Watermarking may be visible or invisible. Invisible
watermarking implies that the presence of the watermark
is barely discernible when the watermarked signal is
displayed. If the watermark cannot be easily removed
from the watermarked signal even after applying
common watermarking attacks then it is referred as
robust embedding. The basic components involved in
robust watermarking are watermark embedding, attack,
and watermark detection [9]. In watermark embedding, a
watermark signal (Text, image or audio etc.) is
constructed and then embedded into an original signal
(Video in context with this paper) to produce the
watermarked signal. Once embedding is done, the
watermarked video can be subjected to various attacks
[10]. During watermark detection, the watermark detector
is given a test signal that may be watermarked, attacked
or not. The watermark detector reports whether the
watermark is present or not on examining the signal at its
input [11].
Video watermarking can be classified categories based
on the method of embedding watermark information bits
in the video. The two categories are: Spatial
watermarking, and transform-domain watermarking. In
spatial watermarking, embedding and detection are done
on spatial pixel values or on the overall video data.
Spatial domain techniques are easier to implement,
however not robust against signal processing operations
like video compression. Transform domain algorithm,
alter spatial pixel values of the video according to pre-
determined transforms. Commonly used are the Fast
Fourier Transform (FFT), Discrete Cosine Transform
(DCT), the Singular Value Decomposition (SVD) and the
Discrete Wavelet Transform (DWT). The transform-
domain watermarking have proved to be more robust and
imperceptible when compared to spatial domain
transforms, as they disperse the watermark in a special
domain of video frame, proving it to be very difficult to
remove [12]. The DWT sections a picture into a lower
determination rough guess picture (LL) and also level
(HL), vertical (LH) and inclining (HH) subtle element
parts. The methodology can then be rehashed to process
numerous “scale” wavelet deterioration. One of the
numerous favourable circumstances over the wavelet
change is that that it is accepted to all the more precisely
model parts of the HVS as contrasted with the FFT or
DCT.
II. IMPORTANT ASPECTS OF VIDEO WATERMARKING
Video watermarking embeds data in the video for the
purpose of identification, annotation and copyright. A
number of video watermarking techniques have been
proposed. These techniques exploit different ways in
order to embed a robust watermark and to maintain
original video fidelity. Conventional encryption
algorithms permit only authorized users to access
encrypted digital data. Once such data are decrypted,
however, there is no way in prohibiting its illegal copying
and distribution [13].
Many algorithms for developing watermarks on
images are extended for videos. But some points need to
be considered during the extensions.
a) Between the frames there exists a huge amount of
intrinsically redundant data.
b) There must be a strong balance between the motion
and the motionless regions
c) Strong concern must be put forth on real time and
streaming video applications.
The following aspects are important for the design of
video watermarking systems.
a) Imperceptibility: The watermark embedding should
cause as little degradation to the host video as possible.
b) Robustness: The watermark must be robust to
common signal processing manipulations and attempts to
remove or impair the watermark.
c) Security: The embedded information must be secure
against tampering.
d) Capacity: The amount of embedded information
must be large enough to uniquely identify the owner of
the video.
Video watermarking is not a standalone technology. It
can be associated with different approaches to achieve a
sophisticated system. This research can be continuous by
applying this new proposed scheme to specific
environment or application and examine its usefulness.
III. SINGULAR VALUE DECOMPOSITION
Singular Value Decomposition (SVD) is a numerical
method for diagonalizing networks in which the changed
area comprises of premise expresses that is ideal in some
sense [8]. The SVD of a N×N network An is
characterized by the operation:
A=U S VT (1)
where U and V Є R N×N are unitary, and S Є R N×N is
an inclining grid. The diagonal entries of S are known as
the peculiar estimations of A and are thought to be
organized in diminishing request σi>σi+1. The sections
of the U grid are known as the left solitary vectors while
the segments of the V network are known as the right
independent vectors of A. Every peculiar worth σi points
out the luminance of a picture layer while the comparing
pair of independent vectors defines the geometry of the
picture [14]-[17].
IV. PROPOSED WORK
We propose a new innovative digital video
watermarking scheme which applies hybrid approach
using Singular Valued Decomposition and Redundant
Wavelet Transform in video. The proposed scheme is
robust against many watermark attacks, as the watermark
is embedded in the frames of video is strategically placed
with complex approach. To enhance the fidelity of the
scheme, key generation and wavelet based key
embedding watermarking scheme is presented. The new
watermarking scheme proposed is based on hybrid model
using singular values from watermark image after
resizing and using singular values of the wavelet
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 570
decomposed frame’s selected layer and also embedding
the watermark key obtained using key generation with
watermarks decomposed orthogonal values and
embedding in the 4-level decomposition of the selected
low energy band of the decomposed band. Experiments
have been performed on this video watermarking scheme
to prove its performance. We compare the results of the
proposed scheme with the existing base scheme with
different parameters and discuss the advantages and the
disadvantages of proposed scheme. The effectiveness of
this scheme is verified through a number of experiments.
A. Embedding Steps
1) Use of secret key, generation and embedding
Digital signature of the orthogonal matrices is a unique
binary string generated through a hashing function. In
addition, the digital signature must be random, so that an
attacker cannot predict them. Watermarking for tamper
detection leads to a similar situation as watermarking
videos. Being a secret key scheme, the same key is used
for both watermark insertion and extraction. Hence, the
key must be transmitted from the owner to the verifier
through a secure channel. Signature bits should be
embedded into high energy region for improved
robustness. Signature should remain robust against wide
range of attacks hence one set of signature bits are
embedded into LL4 and another set is embedded into HH4
band to ensure recovery from at least one of the band. The
algorithm for embedding and extracting the signature is
as follows.
2) Generation of secret code
(i) Add all elements of the column of orthogonal
matrices U and V obtained by SVD decomposition of the
prepared watermark and create 1×N array of values,
where N is the number of rows of the watermark logo.
(ii) Based on the median values of the U & V 1×N
dimension array, map the array values into corresponding
binary codes and obtaining array of the same size.
(iii) By XORing the binary array generates the code
for the given orthogonal matrices U & V.
3) Code embedding
(i) Generate the code of N bits for the U and V
matrices of watermark and the secret key by user and
then applying fuzzy generator to it.
(ii) Using Haar wavelet, decompose the selected layer
of the frame using the image into 4 sub-bands: LL, HL,
LH, and HH. Further decompose LL band to the 4th level.
(iii) Select N random coefficient from LL4 and HH4
band. Convert the integer part into the binary code of L
bits.
(iv) Replace the nth
bit of the coefficient with code bit
and then convert the binary code to its decimal
representation.
(v) Apply the IDWT with modified LL4 and HH4
band coefficients.
4) Watermark embedding algorithm
(i) Select the video to be watermarked and convert it
into frames of RGB24.
(ii) Convert the image frames to double scale.
(iii) Separate the layers of the image and select the
layer or layers to be watermarked.
(iv) Apply 2D wavelet and decompose image layer
into four sub-bands: LL, HL, LH, and HH on selected
layer of the image.
(v) Watermark W is decomposed using SVD and apply
SVD to Diagonal HH band and replace the values of
singular matrix.
W=Uw*Sw*VwT (2)
(vi) Select the other two layers and perform
watermarking on them by following steps VI to VII for
each layer.
(vii) Apply Redundant Wavelet Transform to the
Watermark image and then perform SVD to the HH band
filter image and replace the singular values with the
redundant singular values of the band to obtain second
watermark. Do the step for both blue and green layers.
Wr=Uw*Srw*VwT (3)
(viii) Repeat steps iv to vii for all the frames in the
video.
(ix) Convert all the watermarked frames to uint8
format and convert the frames into video format.
B. Extraction Steps
1) Code extraction
(i) Using Haar wavelet, decompose the selected layer
of the frame using the image into 4 sub-bands: LL, HL,
LH, and HH. Further decompose LL band to the 4th level.
(ii) Select N random coefficient from LL4 and HH4
band. Convert the integer part into the binary code of L
bits.
(iii) Extract the nth
bit from the coefficient to extract
the code.
(iv) Generate the code of N bits for the U and V
matrices of watermark and the secret key by user and
then applying fuzzy generator to it and compare it with
extracted code. If they match, authenticate U and V
matrices and use them in watermark estimation.
Decoder extracts the code and matches it with the
regenerated code for authentication of U & V matrices. If
matching criteria is satisfied, then decoder will continue
estimating watermark.
2) Watermark extraction algorithm
(i) Select the Watermarked Video and convert it into
frames of RGB24.
(ii) Convert the image frames to double scale. Separate
the layers of the image and select the layer or layers to be
watermarked.
(iii) If the Extracted signature is same as the generated
signature the normal extraction takes place.
(iv) Apply 2D wavelet and decompose image layer
into four sub-bands: LL, HL, LH, and HH on selected
layer of the image.
(v) Apply SVD to HH band.
HHw = Uh * Sh * Vht (4)
(vi) Extract the Singular values and recompose the
watermark using inverse SVD on it U and V matrix.
(vii) If the extracted signature does not match the
generated signature, we go for alternate extraction.
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 571
(viii) Selecting the blue and green layers of the image
and apply 2D wavelet and decompose image layer into
four sub-bands: LL, HL, LH, and HH on selected layer of
the image.
(ix) Using redundant wavelet, decompose the
watermark image into four and apply SVD to the HHr
filter image.
(x) Extract Singular values from the HH band obtained
using 2D-DWT in the layers decomposed and apply
inverse SVD to the HHr with the extracted value of
watermark.
HHw=Uw*Swe*Vwt (5)
(xi) Apply inverse redundant wavelet transform to
recompose the watermark with extracted values of the
watermark.
(xii) Compare the normalized co-relation with the
original watermark and plot the results.
(xiii) Compare the PSNR values and structural
similarity of the Original video to the watermarked video
and plot the results.
Fig. 1 and Fig. 2 show the block diagrams for
proposed Embedding and Extraction Process.
Figure 1. Proposed watermarking embedding process
Figure 2. Proposed watermarking extraction process
V. RESULTS
A. PSNR Values
Fig. 3 and Fig. 4 show the PSNR values for Base
scheme and Proposed scheme. Hence, this prompting a
PSNR degradation of around 29.79dB (28th
frame) for
Proposed Scheme and 24.07dB (28th
frame) for base
scheme.
Figure 3. PSNR values for base scheme
Figure 4. PSNR values for proposed scheme
B. Correlation Values
Fig. 5 and Fig. 6 show the Correlation values for Base
scheme and Proposed scheme. Hence, Correlation value
is 0.9179dB (28th
frame) for Proposed scheme and
0.9031dB (28th
frame) for Base scheme.
Figure 5. Correlation values for base scheme
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 572
Figure 6. Correlation values for proposed scheme
C. SSIM Values (Structure Similarity Index)
Fig. 7 and Fig. 8 show the SSIM values for Base
scheme and Proposed scheme. Hence, structural
similarity index increase of 0.9977dB (28th
frame) for
Proposed Scheme and 0.9561dB (28th
frame) for Base
scheme.
Figure 7. SSIM values for base scheme
Figure 8. SSIM values for proposed scheme
D. Extraction of Watermark without Any Attack
Fig. 9 shows retrieved Watermarks for Base Scheme
and Proposed Scheme. This is the normal outcome
without any Attack. Thus, visibility of watermark is more
using Proposed Scheme.
Figure 9. Retrieved watermark without any attack
E. Extraction of Watermark with Crop Attack
Fig. 10 shows retrieved watermark with crop attack
both for base scheme and proposed scheme. Hence,
watermark is more visible using Proposed Scheme and
the watermark is almost corrupt using Base scheme.
Figure 10.
F. Extraction of Watermark with Mean Attack
Fig. 11 shows the outcome with Mean Attack both for
Base Scheme and Proposed Scheme. The Visibility of the
watermark is far better using Proposed scheme.
Figure 11. Retrieved watermark with mean attack
G. Extraction of Watermark with Median Attack
Fig. 12 shows the outcome with Median Attack both
for Base Scheme and Proposed Scheme. The Retrieved
watermark is far better using Proposed Scheme and hence
it is more Robust.
Figure 12. Retrieved watermark with median attack
H. Extraction of Watermark with Noise Attack
Fig. 13 shows the outcome with Noise Attack both for
Base Scheme and Proposed Scheme. Thus, the
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 573
Retrieved watermark with crop attack
watermark is extracted successfully using Proposed
Scheme and watermark is corrupted using Base scheme.
Figure 13. Retrieved watermark with noise attack
I. Extraction of Watermark with Rotation Attack
Fig. 14 shows the outcome with Rotation Attack both
for Base Scheme and Proposed Scheme. Hence, Proposed
watermarking scheme is more robust as its watermark is
clearly visible with rotation attack.
Figure 14. Retrieved watermark with rotation attack
VI. CONCLUSION
The proposed scheme satisfies the requirement of
imperceptibility and robustness for a feasible
watermarking scheme. Moreover, A Blind process is
carried out so it does not need original data at the time of
Extraction and also it does not need any information in
the detection process while other algorithms and yet can
resist majority of the attacks in the process and no
authentication is given as a result tampering detection
cannot be detected. The watermark recovery is better at the
cost of perceptibility of the watermarked video in different
methods. The proposed scheme as discussed is robust
against attacks like Crop, Mean, Median, Noise and
Rotation attacks.
REFERENCES
[1] N. R. Bamane and S. B. Patil, “Comparison & performance analysis of different digital video watermarking techniques,”
International Journal of Scientific & Engineering Research, vol. 4,
no. 1, January 2013. [2] S. Batra and H. K. Khattra, “An improved data transfer technique
using steganography with watermarking and visual cryptography,”
International Journal of Innovative Technology and Exploring Engineering, vol. 3, no. 7, December 2013.
[3] P. V. Powar and S. S. Agrawal, “Design of digital video
watermarking scheme using matlab simulink,” IJRET, vol. 2, no. 5, May 2013.
[4] H. Shojanazeri, W. A. W. Adnan, and S. M. S. Ahmad, “Video
watermarking techniques for copyright protection and content authentication,” International Journal of Computer Information
Systems and Industrial Management Applications, vol. 5, pp. 652-
660, 2013. [5] M. A. Chimanna and S. R. Khot, “Digital video watermarking
techniques for secure multimedia creation and delivery,”
International Journal of Engineering Research and Applications, vol. 3, no. 2, March-April 2013.
[6] S. S. Bedi, R. Ahuja, and H. Agarwal, “Copyright protection using video watermarking based on wavelet transformation in
[7]
[8] the art, contributions and research challenges,” International Journal of Advanced Computer Science and Application, vol. 3,
no. 7, 2012.
[9] M. M. Sathik and S. S. Sujatha, “A novel DWT based invisible watermarking technique for digital images,” International Arab
Journal of e-Technology, vol. 2, no. 3, January 2012.
[10] S. Rungraungsilp, M. Ketcham, P. Surakote, and S. Vongpradhip, “Data hiding method for QR code based on watermark by
comparing DCT with DWT domain,” in Proc. International
Conference on Computer and Communication Technologies, May 26-27, 2012.
[11] J. Madia, K. Dave, V. Sampat, and P. Toprani, “Video
watermarking using dynamic frame selection technique,” in Proc. National Conference on Advancement of Technologies, 2012.
[12] B. Isac and V. Santhi, “A study on digital image and video
watermarking schemes using neural networks,” International Journal of Computer Applications, vol. 12, no. 9, January 2011.
[13] C. Cruz-Ramos, R. Reyes-Reyes, M. Nakano-Miyatake, and H.
Perez-Meana, “A blind video watermarking scheme robust to frame attacks combined with MPEG2 compression,” Journal of
Applied Research and Technology, vol. 8 no. 3, pp. 323-339,
December 2010. [14] L. Rajab, T. Al-Khatib, and A. Al-Haj, “Video watermarking
algorithms using the SVD transform,” European Journal of
Scientific Research, vol. 30, no. 3, pp. 389-401, 2009. [15] G. Bhatnagar and B. Raman, “A new robust reference
watermarking scheme based on DWT-SVD,” Computer Standards
& Interfaces, vol. 31, pp. 1002-1013, 2009. [16] H. H. Mirza, H. D. Thai, Y. Nagata, and Z. Nakao, “Digital video
watermarking based on principal component analysis,” in Proc. Second International Conference on Innovative Computing,
Informastion and Control, 2007.
[17] A. Al-Haj, “Combined DWT-DCT digital image watermarking,” Journal of Computer Science, vol. 3, no. 9, pp. 740-746, 2007.
Sakshi Batra is currently doing her M.Tech
Degree in Electronics and Communication in
Chandigarh Group of Colleges, Landran, Mohali (Punjab). She completed her B.E.
degree in Electronics and Communication from
Lovely Professional University, Jalandhar, India. Her area of interest includes Image
Processing, Networking and Network Security.
Dr. Rajneesh Talwar is presently working as
Principal of Chandigarh Group of Colleges-
COE, Jhanjeri, Punjab, India. He did his PhD in 2010 and M.Tech in 2002 from Thapar
University, Patiala, Punjab, India. He has
worked as Principal of CGC, Landran and Swift Technical Campus, Rajpura. He has been
Vice principal at RIMT Aggrasen Engineering
college, Mandi gobindgarh and Head, Electronics and communication engineering
department at RIMT-IET, Mandi gobindgarh, Punjab India.
Dr. Talwar has a U.S patent “Fiber Optic Point Temperature Sensor” to his credit, twenty + international Journal Publications, presented papers
/participated in more than twelve International conferences and many
national level conference participations. Invited Reviewer of 2014 IEEE Colloquium on Humanities, Science and Engineering Research
(CHUSER 2014), was Reviewer of MAEJO International Journal of
Engineering Science and Technology, Thailand and “Materials and Design”, a ELSEVIER International Journal.
International Journal of Electronics and Electrical Engineering Vol. 4, No. 6, December 2016
©2016 Int. J. Electron. Electr. Eng. 574
multiband,” International Journal of Computer Applications, vol.
66, no. 8, March 2013.S. N. Ahmed, B. Sridhar, and C. Arun, “Robust video
watermarking based on discrete wavelet transform,” International
Journal of Computer Network and Security, vol. 4, no 1, Jan.-Mar.2012.
R. A. Sadek, “SVD based image processing applications: State of