Cryptography Reference
In-Depth Information
where c
′
ij
(n) denotes the modified coe
cient, % denotes the modulo operation
andrthe absolute value of r.
Finally, use the inverse DCT operation to obtain the dual watermarked
image.
In the invisible watermark extraction process, the invisible watermark bit
′
′
W
is extracted from the modified coe
cient c
ij
(n)byusing
′
′
Int
c
ij
(n)/L
%2,
if c
ij
(n) < 0;
Int
′
W
=
(11.21)
c
′
if c
′
1−
ij
(n)/L
%2,
ij
(n)≥0.
11.3.2 Experimental Results
In our experiments, the 512512 256-grayscale Lena image, a 512512 256-
grayscale visible watermark, a 6464 binary invisible watermark are used.
The visible and invisible watermarks are shown in Fig. 11.17 We decompose
the Lena image and the visible watermark into nonoverlapping 88blocks.
The invisible watermark is embedded in the selected AC DCT coe
cients of
midrange frequencies, and the visible watermark is embedded in others. The
dual watermarked image and the corresponding extracted invisible watermark
are shown in Fig. 11.18. To test the robustness of the proposed algorithm, we
made several attacks on the watermarked image. These include JPEG com-
pression, median filtering and Gaussian filtering in Stirmark [31] benchmark
software, together with contrast enhancement and cropping in Adobe Pho-
toshop 5.0 software. Visible watermark removal was done by exploiting the
original visible watermark. The extracted invisible watermarks are shown in
Figs. 11.19(a)(h) and evaluated by normalized Hamming similarity (NHS).
The results show that the proposed algorithm is robust to normal attacks.
The invisible and visible watermarks are independent of each other and the
quality of the watermarked image is well preserved. The invisible watermark
extraction is blind as the original image is not required during the extraction
process.
11.3.3 Conclusion
A novel multipurpose image watermarking technique based on DCT is pre-
sented. The basic idea is to embed the invisible watermark in some frequency
bands, and embed the visible watermark in others. Future work will be fo-
cused on improving the robustness of the invisible watermark to geometric
attacks.
11.4 Image Retrieval Based on a Multipurpose
Watermarking Scheme
Due to the development of computer, multimedia, and network technologies,
two important issues have arisen nowadays. The first problem is that the
