Information Technology Reference
In-Depth Information
the noise dominant ones,
s
is the level or scale used in the wavelet analysis
(
s
= 1 and
s
= 2 for Haar and db4 wavelets, respectively),
is the
standard deviation of the current wavelet coeffi cient kernel in the
s
level,
and 2
−s
is the weighting function [8, 25, 31].
Step 5: Robust criterion to hide data
The image quality of the pro-
cessed images is improved using a criterion based on the median estimator.
The condition
sx
α=σ
s
k
is applied for each kernel of the standard deviation
β
≥
T
, if this condition is satisfi ed then this area or region is considered
noisy and thus can be inserted the information to hide in the respective
wavelet kernel coeffi cients
α
=
σ
s
k
H
of the cover color image [8],
D
,
,oth rw e
β
≥
T
=
⎩
(5)
k
sx
S
k
H
k
()
s
()
s
where
β
=
σ
−
MED
α
and
are the two robust criteria to
β
=
MED
α
s
1
c
s
2
()
s
hide data [26, 27],
MED
α
is the median of the wavelet kernel coeffi-
cients
α
, and
σ
is the standard deviation located at the center of the kernel
s
c
α
. We propose the use of median as a robust estimation of the energy [21,
32] of the wavelet kernel coefficients given by its local standard deviation.
These procedures improve the features of the proposed method to provide
good invisibility, color retention, and
fine detailed preservation of the
processed images [8].
To recover the hidden image, the algorithm is used again but in the step
1 the input changes from the cover color image to the stego-image, and to
repeat the same steps changing the conditions of the step 5 to recover the
hidden data in the following way [8],
s
S
α
<
T
⎧
k
s
D
=
k
H
otherwise
(6)
k
S
,
β
≥
T
⎧
k
x
D
=
(7)
k
H
,
otherwise
k
Other versions of the CLCES algorithm are given using a threshold based
on the median of the standard deviations in the current wavelet kernel co-
efficient in the level
s
changing the threshold of
(
)
∑
∑
−
s
−
s
{}
s
T
=
α
⋅
2
2
T
=
MED
α
to
[8].
s
1
s
s
Search WWH ::
Custom Search