Cryptography Reference
In-Depth Information
visibly watermarked content can be used as in sales promotion or to publi-
cize a reduced-price product. Now, consider that the provider may advertise
on the Internet in order to sell an image that is visibly watermarked. Some-
one wants to buy this image through Internet by using a credit card. Under
the circumstances, three serious problems may be encountered. The first is,
if the customer wants to buy the image without the visible watermark, the
provider or the authorized buyer can automatically remove this from the wa-
termarked image. However, the customer may illicitly resell the one without
visible watermark to others. Secondly, someone may directly download the
visibly watermarked image and sell it to others in a low price. Thirdly, some-
one may illicitly embed the providers visible watermark in his own products
for sales promotion.
In fact however despite the robust properties of a visible watermark, it
can always be partially or fully removed or tampered with using software. To
provide further protection for the image work, this section gives a method
to embed an invisible watermark as a back-up in the host image. This is in
addition to the visible watermark. The two watermarks are embedded in the
image simultaneously and independently. The visible watermark is embedded
in the DC DCT coe
cient and most of the AC DCT coe
cients of each block.
The invisible watermark in the rest. Based on the characteristics of Human
Visual System (HVS), the embedding strength of the visible watermark is
varied in accordance with the underlying content of the host image. We em-
bed the invisible watermark in a number of midrange frequencies. Since the
invisible watermarks inserted in the high frequency components are vulnera-
ble to attack but the low frequency components are perceptually significant.
Alterations to the low frequency components may become visible.
11.3.1 Watermarking Algorithm
In [26], a visible watermarking technique that modifies all the DCT coe
cients
of each host image block using the following equation is proposed:
′
c
ij
(n)=α
n
c
ij
(n)+β
n
w
ij
(n),
(11.14)
where α
n
and β
n
are respectively the scaling factor and the embedding factor
for block n.Thec
ij
(n) are the DCT coe
cients of the block n in the host
image and w
ij
(n) are the DCT coe
cients of block n in the watermark image.
The α
n
and β
n
values are found by using a mathematical model developed by
exploiting the texture sensitivity of the HVS. We can embed more information
in the block where energy tends to be evenly distributed. That is, the AC
DCT coe
cients have small variances. For convenience, the scaling factor α
n
is assumed to be proportional to the variance σ
n
and the embedding factor β
n
to be inversely proportional to σ
n
. The blocks with mid-intensity gray scales
are more sensitive to noise than those of low intensity blocks as well as high
intensity blocks. If the mean gray scale of each image block is denoted as µ
n
