Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
=
= + + + +
Now, two features of eigenvalues and eigenvectors are
described that have been used in the proposed method.
a. If matrix has symmetry hermitian (hermitian is equal
to its own conjugate transpose), then eigenvalues are
real numbers and the corresponding eigenvectors
will be orthogonal. This principle is demonstrated by
Equation (4).
(4)
*
i i
H
i j
i j
A A
x x
=
b. If A is real and symmetric and simultaneously the
results in the previous state eigenvectors are also
real as illustrated in Equation (5).
(5)
* *
*
i i
T
i i
A A
A A x x
= =
= =
It should be noted that matrix A is diagonal and the
eigenvalues are sorted in descending order on main diameter
from left up to right down.
(6)
min
max
0 0 0
0 0 0
0 0 0
0 0 0
=
Where MI
and CI
In order to decrease the complexity of calculating elements
of transformation matrix A, in each rows just the two
variables is considered and the rest elements of the row is set
zero. Symmetrical manner of this row to the center of matrix
is repeated in other row.
For simplify the explanation the structure of matrix A is
shown in Equation (9).
Afterward, by solving each group of equations that yields
from each both symmetrical rows, unknowns in
transformation matrix obtained and desired converter to be
achieved. To clarify the above explanation about the
procedure of how to calculate the variables, an example is
illustrated in Equation (10).
(10)
1 1 2 1
2 1 1
m
m m
aCI a CI MI
a CI aCI MI
+ =
+ =
If the arrays of transformation matrix to be integer then
numbers that appears in eigenvectors belong to one of the
following categories.
(11)
{ }
{ }
0.7071 0.7071 1 0 2 1
0.7071 0.7071 0 2
m k
m k
= +
Thus, constant 100 multiplied in transformation matrix
then floating term of arrays is removed. The error that derived
it is very small and negligible.
(12)
'
(100 ) A round A =
Now, the converter is ready to extract the eigenvalues and
eigenvectors by Equations (3) and (1). After calculating the
eigenvalues and eigenvectors we send cover image,
eigenvectors and eigenvalues separately that way of send will
express in the next section. After sending data, at destination
by received cover image and reproduction matrix A with
eigenvalues and eigenvectors, the secret data will be extracted
from cover image by following Equations.
(13)
' '
2 ( )
2
CI
SI MI =
(14)
' '
MI A CI =
(15)
' 1
0.01 ( ) A x x
=
C. Sending information
If one of three factors , x, CI were not received, secret
data from cover image cannot be extracted. Afterward, each
factor is sent independently, which causes high security in
data sending. In the proposed method, vectors in Equation
(16), (17) which are made from eigenvalues and eigenvectors
will be sent as keys separately. The first vector contains the
eigenvalues that are the elements on the main diagonal of
(16) [ ]
min 2 1 max m
The second is the vector containing the eigenvectors of
transformation matrix A. To create the second vector, matrix
x which contains eigenvectors is swept from upper left to
lower right. If the dimension of A is even. We encounter
along with numbers 0, 0.7071 and -0.071. In order to
decrease the number of required bits for sending data, we
send 1, -1 instead of 0.701 and -0.701 respectively. For
sending zeros, zeros between two non-zero along math are
counted that matrix swept and number of zeros be send.
If there is just one zero, afterward zero will be sent. For
example, the bellow vector is created in Equation (17) for one
matrix x while its dimension is considered 4*4.
(17)
[
]
1 2 1 0 1 1
2 1 1 0 1 2 1
As mentioned above, in receiver with the received cover
image and transformation matrix A by using Equations (13),
(14), and (15) secret image is extracted.
In the proposed method, the cover image is not changed,
therefore the third part is not suspected on existence of secret
data.
III. EXPERIMENTAL RESULTS
The proposed method is executed on four 256*256 images.
Fig.2 shows examples of the data hiding and the data
extracting process. Then three Images Bridge, boat and
pentagon embedded in three cover image Lena, Pepper and
Baboon. In table 1, the proposed method is compared with two
hiding data methods. The PSNR and capacity are term in db
and bits, respectively. It can be seen that the proposed method
has higher capacity than the other two methods, and the PSNR
of the proposed method due to the cover image did not change
is very high.
Fig 1: the experimental result on four images
TABLE I: A comparison of proposed method and 2-LSB and 3-LSB on capacity and PSNR
Cover
Image
Secret
image
2-bit LSB
Capacity PSNR stego
Image
3-bit LSB
Capacity PSNR stego
Image
Proposed method
Capacity PSNR stego PSNR Extracted
Image Image
Baboon
Lena
Peppers
Bridge
Boat
Pentagon
Bridge
Boat
Pentagon
Bridge
Boat
Pentagon
524,288 44.37
524,288 44.52
524,288 44.51
524,288 44.37
524,288 44.51
524,288 44.52
524,288 44.39
524,288 44.53
524,288 44.54
786,432 37.16
786,432 37.17
786,432 37.10
786,432 37.03
786,432 37.05
786,432 36.97
786,432 37.05
786,432 37.09
786,432 37.02
2,097,152 Inf 48.38
2,097,152 Inf 48.61
2,097,152 Inf 48.48
2,097,152 Inf 47.92
2,097,152 Inf 47.86
2,097,152 Inf 48.01
2,097,152 Inf 48.53
2,097,152 Inf 48.63
2,097,152 Inf 48.51
IV. CONCLUSION
In this paper, a new method based on the fundamental
concepts of matrix eigenvalues and eigenvectors is presented.
This method introduced a symmetric transformation matrix
by the properties of the matrix. The key is provided by using
its and send to the receiver separately. In destination, secret
data is extracted by having the cover image and the secret
key. It should be mentioned that if one of them is not
received, the secret data cannot be extracted.
This method has high security because the cover image
was not changed through the hiding process and the secret
key and cover image send separately. Also, this method has
higher data capacity than previous works because information
is not embedded in cover image. In other words, information
is inserted on the secret key.
The main drawback of proposed method is change of
extracted image in comparison with secret image. It should be
mentioned that it is negligible against capacity and robustness
of proposed method.
REFERENCES
[1] Y. M. Behbahani, P. Ghayour, and A. H. Farzaneh, Eigenvalue
Steganography based on eigen characteristics of quantized DCT
matrices, Information Technology and Multimedia(ICIM).14-16
NOV.2011.
[2] Z. Ni, Y.Q. Shi, and N. Ansari, W.Su, Reversibale data hiding, circuits
and system for video technology, IEEE Trans,No:16, pp:354-
362.2006.
[3] V. Sachnev, H. J. Kim, J. Nam, S. Suresh, Y. Q. Shi, Reversibel
watermarking algorithm using sorting and prediction, IEEE Trans.
Circuits and Systems for video Technology Vol. 19, No. 7, pp. 989-999,
2009.
[4] Y. Wang, P. Moulin, Steganalysis of block-DCT image
steganography, IEEE workshop on Statistical Signal Processing, IEEE
Computer Society Press, Los Alamitos, 2003.
[5] J. Silman, Steganography and Steganalysis: An Overview, SANS
Institute, 2001.
[6] Y. Q. Shi, H. Sun, "Image and Video Compression for Multimedia
Engineering, CRC Press, Boca Raton London New York Washing,
D.C. 2001.
[7] C. K. Chan, L. M. Chan, Hiding data in image by simple LSB
substitution,Pattern Recognition,Vol. 37,pp. 469-467, 2004.
[8] A. Ker, Improved detection of LSB steganography in grayscale
image, Lecture Notes in Computer Science ,pp. 97 -115, 2005.
[9] J. Fridrich and M. Goljan, Digital image steganography using
stochastic modulation, SPIE Symposium on Electronic Imaging, San
Jose, CA,, 2003.
[10] C. Cachin, An Information-Theoretic Model for Steganography,,
Proceedings of 2
nd
Workshop on Information Hiding, MIT Laboratory
for Computer Science, May 1998.