Beruflich Dokumente
Kultur Dokumente
Video
I.
Objd
Steganographic
System,
INTRODUCTION
TAlH3
Messagk N
Stereoscopic Pairs-
Video Objects
Extraction Module
Im rint
Msssags-lmagcN(OU-
Selection
Video6batN
L
Estimation
Module
11.
111.
Figure 2: Unsupe~sedemaction of the fmt cova video object. (a) Original left channel. @) Original right channel. (c) Color
segments mask. (d) Depth segments mask. (e) Projection of the color segments onto the depth segments mask. (0Fusion of the
color segments belonging to the same depth segment. (s) Extracted foreground video object.
(I)
Forj=l toMpl
1fx7/i.i)ZT,
I* M~zxNp,isthesizeofsubbandLHz'l
~~
i=l j - I
p-1
PI!
A.
Ft+l
End If
End If
End Forj
End Fori
Afterwards swnmatiou of the coefficientso f QSWT[i] for
i-0 to I is calculated and if the final message-image is of size
axb then the top axb QSWTs (according lo summation) we
selected for embedding the message. For this reason initially
the gray levels of the fmal message-image are sorted in
descending order prcducbg a gay-levels mahix. Then for
i=l to axb the coefficients +CO of the gray-levelsmatrix are
embedded as follows:
X ' k i J F X2(iJI*(l+CZX wfil)),
(4)
where x2(ij)ELHz, c2 is a scaling ,constant that balances
unobsmcmess and robustness and x&' is a coefficient of
the LH2 subband ofthe stego-object. This non-linear insertion
.,
..
Figure 3: Unsupavised extraction of the second mver video object. (a) Original lefl channel. @) Extracted foreground video object.
procedure is similar to [ 141 and adapts the message to the
energy of each wavelet coefficient. Thereby when x&> is
small, the embedded message energy is also small to avoid
artifacts, while when xt(ij> is large the embedded message
energy is increased for robustness, Similarly for the
coefficients of subband LH, we have:
..
111
E.
MrrsogeRecovery
rm
Y=DWTo()
\-,
U'=DWf&)
(7)
Step2 Using the size of message-image (axb). the
embedded posihons are delccted by ~&OWI& the hidmg
process deseribed in rubsenion 3.1. Then the coefftc~cntsof
subband LH?(LHI)of Y are subtracted fmm the cocfbcnts
of subband LH, (LH,)of Y and the result IS scaled down by
thc value of coefficient of Lfl> (LII,) of Y, multiplied hy e,
(E,).
Fori=l t o a x b
w:'qx;z'W+(X
x,q,
:'I-
(X,"l*C>)
X,"II (X,%,)
(8)
(9)
message-imagc.
St~p4:The original message is recovered by decrypting
the cnciphcmd mcssagc. impinled onto thc message-image.
FIEYRCNALMU
FASDFJFDGIGU
YEWCIBDFGRAI
(4
YWOURFOWFSL
FllWHAFQWFAP
STFASGTNQIERP
(4
(e)
Figure 4:Final content generation (a) Message-image of the fmt case. @) Steg+object of the first case ( P S M . 42.3 dB). (c) Messageimage of the second case. (d) Stego+bject of the second case (PSNR: 445 dEl). (e) Synthesized fmal content using the two stegoubjects
and a backgroundvideo object.
BERs.
Table 11: Message retrieval results for the stego-object of Figure 4(d), under different combinations of compression ratios and
BERs.
The final content image i s depicted in Figure 4(e). This
image was then tested under different JPEG compression
ratios and Various Bit Error Rates (BERs). In particular
compression ratios of 2.6.and 5.1 were used to decrease the
mount of data. Then lossy transmission simulations were
performed for 3 different BERs of 3x104, IxIlT and 3 ~ 1 0 . ~
for the compressed fmal content image, considering that
typical average BERs for cellular mobile radio channels are
between IO and ID- [16]. Results of the message-images
for the fmt and second video objects are given in
Tables I and II respectively. In these tables PSNRs have been
ACKNOWLEDGEMENTS
REFERENCES
[5] E.Milbrand~hnp://members.~pod.comlstegano~
aphy/stego.html, September 2001.
[6] N.F. Johnson, and S. lajodia, Steganography:
Seeing the Unseen, IEEE Computer, February
1998, pp. 26-34.
[7] A. Westfield, and A. Pfimann, Attacks on
Steganographic Systems, Proc. 3d Information
Hiding Workhop, Dresden, Germany, Sept.
28-Oct. I , 1999, pp. 61-75.
[SI J. Fridrich, R Du, and L, Meng, Steganalysis of
LSB Encoding in Color Images, in Proceedings
of the IEEE International Conference o n
Multimedia and Expo 2000, Jul.-Aug. 2000, N.Y.,
USA.
[9] L. M. Marvel, C. G. Boncelet, Jr., and C. T.
Retter, Spread Spectrum Image Steganography,
IEEE Transactions on Image Processing, vol. 8,
no. 8, pp. 1075-1083, August 1999.
[IO] S. Areepongsa, Y.F. Syed, N. Kaewkamnerd, and
K.R. b o , Steganography For a Low Bit-Rate
Wavelet Based Image Coder, in Proceedings of
the IEEE International Conference on Image
Processing 2000, Sept. 2000, Vancouver, Canada.
[ 1 I] Douglas Stinson, Cryptography: Theory and
Practice, CRC Press, Inc., 1995.
[I21 M.-S. Hsieh, D.-C. Tseng, and Y.-H.Huang,
Hiding
Digital
Watermarks
Using
Multiresolution Wavelet Transform, IEEE.
Trans. Industrial Electronics, vol. 48, no. 5,
pp.875-882, O c t 2001.
[13] J. M. Shapiro, Embedded Image Coding Using
Zeromes of Wavelet Coefficients, IEEE Trans.
Signal Processing, ~01.41, pp.3445-3462, Dec.
1993.
[I41 X. Wu, W. Zhu, Z. Xiong, and Y.-Q. Zbang,
Object-based multiresolution watermarking of
images and video, in Proceedings IEEE
International Symposium on Circuits and Systems,
Geneva, Switzerland, May 28-31,2000.
[I51 A. D. Doulamis, N. D. Doulamis, K. S. Ntalianis
and S. D. Kollias, Efficient Unsupervised
Content-Based Segmentation in Stereoscopic
Video Sequence, in International Journal on
Artijicial Intelligence Tools, vol 8, n o 6 2000.
[I61 V. Weerackody, C. Podilchuk, and A. Estrella,
Transmission of PEG-Coded Images over
Wireless Channels, Bell Labs Technical Journal,
Autumn 1996.