National Conference on Recent Trends in Engineering & Technology

Invisible Digital Video Watermarking Using 4-level

DWT
Snehal V. Patel

Prof. Arvind R. Yadav

Electronics and Communication Department

Parul Institute of Engineering and Technology
Vadodara, India
snehalpatel01@gmail.com

Electronics and Communication Department

Parul Institute of Engineering and Technology
Vadodara, India
aryme@rediffmail.com

AbstractIn this paper, a algorithm based video watermarking

scheme in the discrete wavelet transform (DWT) domain is
proposed. Scene change analysis is first conducted to decompose
video into different scenes. Each frame of the video is
transformed to wavelet domain by DWT. The watermark image
is decomposed into 8-bit planes, scrambled and embedded into
the mid-frequency DWT coefficients. The quality of the
watermarked video is enhanced by GA. Experimental results
demonstrate that it is robust to common attacks in video
watermarking such as frame dropping, frame averaging additive
noise and lossy compression. (Abstract)
Keywords-Abrupt scene change, peak signal noise ratio
(PSNR), video watermarking, discrete wavelet transform,
robustness. (key words)

I.
INTRODUCTION
Digital watermarking is an effective way to protect
copyright of multimedia data even after its transmission.
Watermarking is a concept of embedding a special pattern,
watermark, into a multimedia document so that a given piece
of copyright information is permanently tied to the data. This
information can later prove the ownership, identify a
misappropriating person, trace the marked documents
dissemination through the network, or simply inform users
about the rights-holder or the permitted use of the data. Most of
the proposed video watermarking schemes are based on the
techniques of image watermarking and directly applied to raw
video or compressed video.Video watermarking introduces
some issues which is not present in image watermarking. Due
to a large amount of data and inherent redundancies between
frames, video signals are highly susceptible to attacks, such as
frame averaging, frame dropping, lossy compression and
statistical analysis [1]. However, the currently available
algorithms do not solve these problems effectively. Since the
goals of watermarking, such as robustness, transparency and
capacity, are usually conflicting, GA can be utilized to solve
the optimization problem. Video watermarking schemes must
not use the original video during watermark detection as the
video usually is in very large size and it is inconvenient to store
it twice. We propose a new watermarking scheme to overcome
these problems. The watermarking process overview is
depicted in Fig. 1. Video sequence is pre-processed by scene
change detection and DWT [3].

13-14 May 2011

Fig.1 Overview of watermarking process

B.V.M. Engineering College, V.V.Nagar,Gujarat,India

National Conference on Recent Trends in Engineering & Technology

First of all original color video is stored in one array then it
is converted in to frames. Then scene change detection
algorithm is applied on to the consecutive frames. If scene
change found we have to change the watermark image else
apply the same watermark within one scene of the video.
Decompose video frame and watermark in to three different
components. Apply the 4 level DWT on each component of
video frame. Generate the secret key and add in to watermark
data because watermark image bits are not enough for
embedding in to original image. Apply the embedding
algorithm for watermark embedding and store the watermarked
data in to one array. Apply the 4 level inverses DWT to
watermarked video frames. Repeat the entire step up to number
of frames. Combine R.G.B. component of each watermarked
video frames. So, we get the watermarked video.
II.

WATERMARK IMAGE PRE-PROCESSING

A watermark image is pre-processed and embedded into

different scenes, so that it can resist a number of attacks toward
the video. It is firstly decomposed into binary image. The
watermark binary color image is decomposed in to three
components (Red, Green and Blue) and these three components
are applied to the secret key generation algorithm. Finally three
watermark images are generated which will be embedded in
three different components of original color video frame which
shown in fig.2.
Length of generated watermark image = (length of original
watermark image) + (length of secret key)
To further increase the security of the watermark, it is
scrambled using Arnold transform, which is in fact an iterative
process to move the pixel position. It is defined as follows:
1 1
X
X
=
mod N
1 2
y
Y

range coefficients of the watermark are modified according to

the watermark. The basic idea is that the human eyes are
sensitive to the low frequency noise and the quantization step
of lossy compression may discard the high frequency
components. Therefore, the reasonable trade-off is to embed
the watermark into the middle-frequency range of the video
frames. Each video frame is transformed to the wavelet domain
with 4 levels. Haar wavelet is used for simplicity, and only the
LH1, HL1, LH2 and HL2, LH3, HL3, LH4, HL4 coefficients
are embedded with scrambled watermark. Figure 4 shows the
result of the Lena image after a four-level 2-D DWT.
In addition, with such a scheme, it is not possible to add
more watermark energy at a particular frequency, in which the
image energy is high, in order to improve robustness. LL band
contains the most of energy of the image, so we apply the
watermark in mid frequency bands, it will not creates some
artifacts in the image. And invisibility of the watermark can be
increased. Here embedding process shown in fig.6.

Fig.2 Overview of watermark preprocess

where (x, y) is the pixel position, and (x', y') are the new
position after Arnold transform. N is the pixel width or height.
III.

PREPARE YOUR PAPER BEFORE STYLING

A. Scene Change Detection

Since applying a fixed image watermark to each frame in
the video is hard to maintain statistical and perceptual
invisibility and applying independent watermarks to each
frame cannot resist statistical comparison or average. The new
watermarking scheme is based on scene change detection,
which is to decompose video sequence into different video
scenes. Flowchart of scene change detection is shown in fig. 3.
Independent watermarks are employed for successive but
different scenes and an identical watermark is embedded within
each motionless scene. Histogram analysis is utilized for scene
change detection. If the histogram difference of two scenes is
greater than the threshold, we consider there is a scene change
[5].
B. DWT
The luminance components of the input video frames are
transformed to frequency domain and the middle-frequency

13-14 May 2011

Fig.3 Flow chart for abrupt change in video transition

B.V.M. Engineering College, V.V.Nagar,Gujarat,India

National Conference on Recent Trends in Engineering & Technology

Fig.4 Lena image for Testing

Fig.5 Lena image after a four-level 2-D DWT

Fig. 7 Flow chart for watermark embedding

Fig-6 embedding process

IV.

WATERMARK EMBEDDING AND DETECTION

A. Watermark Embedding
Figure 7 Shows how to embed the Watermark and secret
key in to the video frames here we apply the watermark in to
only mid frequency band of the image because of the effect of
the human visual system. The power present in the frequency
bands varies greatly from image to image. If the image energy
in a particular band is very low and the watermark energy in
that band is high, then some artifacts are created in the image,
since the watermark energy is too strong relative to the image.
Figure 7 shows the Flow chart of embedding algorithm to
embed the watermark in to color video. First of all convert the
video in to frames and then frame to image. Decompose each

13-14 May 2011

image in to three color components (R.G.B). Apply 4-level

DWT to each component. Here we apply watermark in to HL
and LH bands (mid frequency bands) of each level so convert
each pixel value in to binary. To increase the robustness we
apply the watermark in the 5th bit of every pixel. If 5th bit is
equal to 1 then SC( i , j ,k) is equal to 2 else SC( i , j ,k) is equal
to 1, where SC is the array to store the information for original
video Extraction and (i,j) is the pixel position of particular
band and k is the frame number. Then if W( i , j) =1 then 5th
bit of pixel is equal to 1 else it is 0.we start to embed the
watermark from HL4 (4th level mid frequency band) and then
sequence in to LH4, HL3, LH3, HL2, LH2, HL1 and LH1. At
last store the watermarked video frame data in to array sc(i, j,
k) for watermark Extraction before applying inverse DWT.
B. Watermark Detection
The extraction process requires the key used for selecting
the frames, the wavelet transform filter, and the channel I
which the watermark is inserted. Due to the act of requantizing, a threshold region T should be defined in order to
detect the existence of the video watermarking.
Figure 8 shows the Flow chart of extraction algorithm to
extract the watermark and secret key from the watermarked
video. First Load the watermarked data stored in array sc(i,j,k)
where (i,j) is the pixel position in particular frequency band

B.V.M. Engineering College, V.V.Nagar,Gujarat,India

National Conference on Recent Trends in Engineering & Technology

Fig.8 Flow chart for watermark and secret key Extraction

k is the frame number. Lets convert each pixel of mid

frequency band in to binary. Check if 5th bit of pixel value is
1 then WR (i,j)==1 ,where WR is the recovered watermark
and secret key matrix. If 5th bit of pixel value is 0 then
WR(i,j)==0. Decompose watermark data and secret key from
the matrix WR. Apply these algorithms to three different
watermarked images so we can get three different watermark
data, and then combine these data to retrieve the watermark
image.
Figure 10 shows the Flow chart of extraction algorithm to
extract the original video from the watermarked video. Load
the watermarked data stored in array sc(i,j,k) and stored array
SC(i,j,k), where SC contain the original value of the 5th bit
before watermarking. Lets convert each pixel of mid
frequency band in to binary. Check if SC(i,j,k) is 2 then 5th
bit of pixel sc(i,j,k) is 1 else it is 0. Convert each binary
pixel value in to decimal. Combine three array sc(i,j,k) for
three images R,G,B so the original video will retrieved.

13-14 May 2011

Fig.9 Flow chart for original colour video Extraction

I.

SIMULATION RESULTS & DISSUCTION

Fig 10- original video frames

B.V.M. Engineering College, V.V.Nagar,Gujarat,India

National Conference on Recent Trends in Engineering & Technology

Fig 11-watermarked video frames

Fig.14 PSNR between original and watermarked video frames

Fig 12 - Recovered watermark image

Fig.15 Normalized Correlation between original and recovered watermark

images

In order to evaluate the robustness of watermarking

algorithm, the comparability between original watermark w and
detected watermark w* is calculated with the formula given in
Eq.

Fig 13 - Recovered original video frames from watermarked video

In order to evaluate the quality of image, we use parameter

peak value signal-to-noise ratio (PSNR) [6].

NC value is 1 for all images; This shows the 100% recovery

of watermark images. The PSNR value is between 33 to 41.50
db. The PSNR value shows that the algorithm keeps the quality
of the image and invisibility of embedded watermark without
any attacks.
II.

where N is the size of image, f (i, j ), f ' (i, j ) is the pixel

gray value of host image and pending detection image
respectively. Bigger the value of PSNR, better the quality of
image [6].

13-14 May 2011

CONCLUSION

In this paper I propose a scene-based watermarking

scheme. The scheme is robust against various attacks because
we does not require original video as well as watermarked
video for original video and watermark video recovery.
Experiment has been done on these novel video watermarking
schemes to test an show its performance. The robustness of our
approach is demonstrated using the calculation of NC.

B.V.M. Engineering College, V.V.Nagar,Gujarat,India

National Conference on Recent Trends in Engineering & Technology

REFERENCES

[4].

[1]. P.W. Chan and M. Lyu, A DWT-based Digital Video Watermarking

Scheme with Error Correcting Code, Proceedings Fifth International
Conference on Information and Communications Security (ICICS2003),
Lecture Notes in Computer Science, Springer, Vol. 2836, pp. 202-213,
Huhehaote City, Inner-Mongolia, China, Oct. 10-13, 2003.
[2]. P.W. Chan, M.R. Lyu, R. Chin, Copyright Protection on the Web: A
Hybrid Digital Video Watermarking Scheme, Poster Proceedings 13th
International World Wide Web Conference (WWW2004), pp. 354-355,
New York, May 17- 22, 2004.
[3]. N. Checcacci, M. Barni, F. Bartolini, and S. Basagni, Robust video
watermarking for wireless multimedia communications , Proceedings
IEEEE Wireless Communications and Networking Confernce 2000,
WCNC. 2000, Vol. 3, pp. 1530-1535, 2000.

13-14 May 2011

[5].
[6].

F. Duan, I. King, L. Xu, and L. Chan, Intra-block max min algorithm

for embedding robust digital watermark into BIBLIOGRAPHY 128
images, In Horace H.S. Ip and Arnold W.M. Smeulders, editors,
Proceedings of the IAPR International Workshop on Multimedia
Information Analysis and Retrieval, MINAR 98, Lecture Notes in
Computer Science, Vol. 1464, pp. 255-264, Berlin Heidelberg,
Germany, 1998. Springer- Verlag.
Priyadarshinee Adhikari, Neeta Gargote, Jyothi Digge, and B.G.
Hogade,Abrupt Scene Change Detection,World Academy of Science,
Engineering and Technology 42 2008.
Na Li, Xiaoshi Zheng1 , Yanling Zhao, Huimin Wu, Shifeng Li,
Robust Algorithm of Digital Image Watermarking Based on Discrete
Wavelet Transform, International Symposium on Electronic Commerce
and Security.

B.V.M. Engineering College, V.V.Nagar,Gujarat,India