Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more ➡
Download
Standard view
Full view
of .
Add note
Save to My Library
Sync to mobile
Look up keyword
Like this
1Activity
×
0 of .
Results for:
No results containing your search query
P. 1
Paper 7

Paper 7

Ratings: (0)|Views: 244|Likes:
Wisdom Based Computing, Volume 1, Issue 2, August 2011
Wisdom Based Computing, Volume 1, Issue 2, August 2011

More info:

categoriesTypes, Research, Science
Published by: wisdombasedcomputing on Jun 22, 2011
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See More
See less

06/22/2011

pdf

text

original

 
 International Journal of Wisdom Based Computing, Vol. 1 (2), August 2011 7
 Digital Video Watermarking using Discrete Wavelet Transform and Principal Component Analysis
Sanjana Sinha, Prajnat Bardhan, Swarnali Pramanick, Ankul Jagatramka, Dipak K. Kole, Aruna Chakraborty.Department of Computer Science & EngineeringSt Thomas’ College of Engineering and TechnologyKolkata, India.{aruna.stcet@gmail.com, dipak.kole@gmail.com, sanjanasinha89@gmail.com}
 Abstract
- Due to the extensive use of digital mediaapplications, multimedia security and copyright protectionhas gained tremendous importance. Digital Watermarkingis a technology used for the copyright protection of digitalapplications. In this paper, a comprehensive approach forwatermarking digital video is introduced. We propose ahybrid digital video watermarking scheme based on DiscreteWavelet Transform (DWT) and Principal ComponentAnalysis (PCA). PCA helps in reducing correlation amongthe wavelet coefficients obtained from waveletdecomposition of each video frame thereby dispersing thewatermark bits into the uncorrelated coefficients. The videoframes are first decomposed using DWT and the binarywatermark is embedded in the principal components of thelow frequency wavelet coefficients. The imperceptible highbit rate watermark embedded is robust against variousattacks that can be carried out on the watermarked video,such as filtering, contrast adjustment, noise addition andgeometric attacks.
 Keywords:- Digital video; binary watermark; DiscreteWavelet Transform; Principal Component Analysis.
I.
 
I
 NTRODUCTION
 The popularity of digital video based applications [1]is accompanied by the need for copyright protection to prevent illicit copying and distribution of digital video.Copyright protection inserts authentication data such asownership information and logo in the digital mediawithout affecting its perceptual quality. In case of anydispute, authentication data is extracted from the mediaand can be used as an authoritative proof to prove theownership. As a method of copyright protection, digitalvideo watermarking [2, 3] has recently emerged as asignificant field of interest and a very active area of research. Watermarking is the process that embeds datacalled a watermark or digital signature into a multimediaobject such that watermark can be detected or extractedlater to make an assertion about the object. The object may be an image or audio or video. For the purpose of copyright protection digital watermarking techniques mustmeet the criteria of imperceptibility as well as robustnessagainst all attacks [4-6] for removal of the watermark.Many digital watermarking schemes have been proposed for still images and videos [7]. Most of themoperate on uncompressed videos [8-10], while othersembed watermarks directly into compressed videos [8,11]. The work on video specific watermarking can befurther found in [12-15]. Video watermarking introduces anumber of issues not present in image watermarking. Dueto inherent redundancy between video frames, videosignals are highly susceptible to attacks such as frameaveraging, frame dropping, frame swapping and statisticalanalysis.Video watermarking approaches can be classified intotwo main categories based on the method of hidingwatermark bits in the host video. The two categories are:Spatial domain watermarking where embedding anddetection of watermark are performed by directlymanipulating the pixel intensity values of the video frame.Transform domain [16-18] techniques, on the other hand,alter spatial pixel values of the host video according to a pre-determined transform and are more robust than spatialdomain techniques since they disperse the watermark inthe spatial domain of the video frame making it difficult toremove the watermark through malicious attacks likecropping, scaling, rotations and geometrical attacks. Thecommonly used transform domain techniques are DiscreteFourier Transform (DFT), the Discrete Cosine Transform(DCT), and the Discrete Wavelet Transform (DWT).In this paper, we propose an imperceptible and robustvideo watermarking algorithm based on Discrete WaveletTransform (DWT) [19-25] and Principal ComponentAnalysis (PCA) [26-28]. DWT is more computationallyefficient than other transform methods like DFT and DCT.Due to its excellent spatio-frequency localization properties, the DWT is very suitable to identify areas inthe host video frame where a watermark can be embeddedimperceptibly. It is known that even after thedecomposition of the video frame using the wavelettransformation there exist some amount of correlation between the wavelet coefficients. PCA is basically used tohybridize the algorithm as it has the inherent property of removing the correlation amongst the data i.e. the waveletcoefficients and it helps in distributing the watermark bitsover the sub-band used for embedding thus resulting in amore robust watermarking scheme that is resistant toalmost all possible attacks. The watermark is embeddedinto the luminance component of the extracted frames as itis less sensitive to the human visual system (HVS).The paper is organized as follows. Section II containsthe watermarking scheme. Section III contains theexperimental results and finally Section IV gives theconclusion.
 
 International Journal of Wisdom Based Computing, Vol. 1 (2), August 2011 8II.
 
W
ATERMARKING
S
CHEME
 The watermarking algorithm basically utilizes twomathematical techniques: DWT and PCA. Thesignificance of using these techniques in watermarking has been explained first.
 A.
 
 Discrete Wavelet Transform
The Discrete Wavelet Transform (DWT) is used in awide variety of signal processing applications. 2-Ddiscrete wavelet transform (DWT) decomposes an imageor a video frame into sub-images, 3 details and 1approximation. The approximation sub-image resemblesthe original on 1/4 the scale of the original. The 2-D DWT(Fig. 1) is an application of the 1-D DWT in both thehorizontal and the vertical directions. DWT separates thefrequency band of an image into a lower resolutionapproximation sub-band (LL) as well as horizontal (HL),vertical (LH) and diagonal (HH) detail components.Embedding the watermark in low frequencies obtained bywavelet decomposition increases the robustness withrespect to attacks that have low pass characteristics likefiltering, lossy compression and geometric distortionswhile making the scheme more sensitive to contrastadjustment, gamma correction, and histogramequalization. Since the HVS is less sensitive to highfrequencies, embedding the watermark in high frequencysub-bands makes the watermark more imperceptible whileembedding in low frequencies makes it more robustagainst a variety of attacks.
Figure 1. DWT subbands.
 B.
 
Principal Component Analysis
Principal component analysis (PCA) is a mathematical procedure that uses an orthogonal transformation toconvert a set of observations of possibly correlatedvariables into a set of values of uncorrelated variablescalled principal components. The number of principalcomponents is less than or equal to the number of originalvariables. PCA is a method of identifying patterns in data,and expressing the data in such a way so as to highlighttheir similarities and differences. Since patterns in datacan be hard to find in data of high dimension, where theadvantage of graphical representation is not available,PCA is a powerful tool for analyzing data. The other mainadvantage of PCA is that once these patterns in the datahave been identified, the data can be compressed byreducing the number of dimensions, without much loss of information. It plots the data into a new coordinate systemwhere the data with maximum covariance are plottedtogether and is known as the first principal component.Similarly, there are the second and third principalcomponents and so on. The maximum energyconcentration lies in the first principal component.The following block diagram (Fig.2) shows theembedding and extraction procedure of the watermark. Inthe proposed method the binary watermark is embeddedinto each of the video frames by the decomposition of theframes into DWT sub bands followed by the application of  block based PCA on the sub-blocks of the low frequencysub-band. The watermark is embedded into the principalcomponents of the sub-blocks. The extracted watermark isobtained through a similar procedure.
LL, HHLL, HH
Figure 2. Block Diagram of Watermarking
Block basedPCAWatermark Inverse PCAIDWTWatermarkedVideoDWTBlock based PCABlock based PCA of original video frameExtracted Watermark Original videoframeApply DWT
 
 International Journal of Wisdom Based Computing, Vol. 1 (2), August 2011 9
C.
 
 Algorithms for watermarking using DWT AND PCA
Algorithm 1:
a)
 
 Embedding Procedure
Step 1: Convert the n × n binary watermark logo into avector W = { w
1
, w
2
, ……, w
n × n
} of ‘0’s and ‘1’s.Step 2: Divide the video (
2N × 2N
) into distinct frames.Step 3: Convert each frame from RGB to YUV colour format.Step 4: Apply 1-level DWT to the luminance (Ycomponent) of each video frame to obtain four sub-bandsLL, LH , HL and HH of size
 N x N
.Step 5: Divide the LL sub-band into k non-overlappingsub-blocks each of dimension n × n (of the same size asthe watermark logo).Step 6: The watermark bits are embedded with strength
α
 into each sub-block by first obtaining the principalcomponent scores by Algorithm 2. The embedding iscarried out as equation 1.
scorescore
ii
α 
+=
'
(1)where
i
score
represents the principal component matrixof the i
th
sub-block.Step 7: Apply inverse PCA on the modified PCAcomponents of the sub-blocks of the LL sub-band toobtain the modified wavelet coefficients.Step 8: Apply inverse DWT to obtain the watermarkedluminance component of the frame. Then convert thevideo frame back to its RGB components.
b)
 
 Extraction Procedure
Step 1: Divide the watermarked (and possibly attacked)video into distinct frames and convert them from RGB toYUV format.Step 2: Choose the luminance (Y) component of a frameand apply the DWT to decompose the Y component intothe four sub-bands LL , HL , LH , and HH of size N×N.Step 3: Divide the LL sub-band into n × n non-overlapping sub-blocks.Step 4: Apply PCA to each block in the chosen subbandLL by using Algorithm 2.Step 5: From the LL sub-band, the watermark bits areextracted from the principal components of each sub-block as in equation 2.
α 
)(
''
iii
scorescore
=
(2)where
'
i
is the watermark extracted from the i
th
sub- block.Algorithm 2:The LL sub-band coefficients are transformed into a newcoordinate set by calculating the principal components of each sub-block (size n x n).Step 1: Each sub-block is converted into a row vector 
i
 D
 with n2 elements (i=1,2…k ).Step 2: Compute the mean
i
µ 
and standard deviation
i
σ 
 of the elements of vector 
i
 D
.Step 3: Compute
i
 Z 
according to the following equation
iiii
 D Z 
σ µ 
)(
=
(3)Here
i
 Z 
represents a centered, scaled version of 
i
 D
, of the same size as that of 
i
 D
.Step 4: Carry out principal component analysis on
i
 Z 
 (size 1 x n
2
) to obtain the principal component coefficientmatrix coeffi (size n
2
× n
2
).Step 5: Calculate vector 
i
score
as
iii
coeff  Z score
×=
(4)where
i
score
represents the principal component scoresof the i
th
sub-block.III.
 
E
XPERIMENTAL
ESULTS
 The proposed algorithm is applied to a sample videosequence ‘Lake.wmv’ using a 32 × 32 watermark logo.The grayscale watermark is converted to binary beforeembedding. Fig. 3(a) and 3(b) show the original and thewatermarked video frames respectively. Fig. 4(a) is theembedded watermark and Fig. 4(b) is the extracted binarywatermark image.The performance of the algorithm has been measuredin terms of its imperceptibility and robustness against the possible attacks like noise addition, filtering, geometricattacks etc.
(a)

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->