M2USIC 2004 TS4B-3

Real Time DCT-based Multi-Resolution Watermarking Technique

Dr. Fazal-ur-Rehman, Zahid Iqbal, Shaukat Iqbal, Muhammad Zafar Javed Shahid
Ghulam Ishaq Khan Institute of Engineering and Technology Pakistan
Email: [email protected], [email protected], [email protected], [email protected]

Abstract be difficult to be removed, even after various signal

processing techniques. Thus the goal of a label is to
In digital world, the expansion of digital networks always remain present in the data. A watermarking
allows extensive access on, and reuse of, visual algorithm consists of three parts: the watermark, the
material. But at the same time, it has created embedder and the detector. Each owner has a unique
problems, which include unauthorized taping, watermark. The embedding algorithm incorporates the
reading, manipulating or removing of data, which watermark into the image. The detector algorithm
might lead to financial loss or legal problems of the authenticates the image, determining both the owner
producers and creators. Thus, designers, producers and the integrity of the image
and publishers of digital data are seeking technical
solutions to the problems associated with copyright
protection of multimedia data. Digital watermarking 2 ROBUST WATERMARKING
is the technology to prove copyright, monitor the ALGORITHM
usage of the copyrighted multimedia data.
Embedding of unique customer identification into There are two method of embedding watermark in
data is used to identify illegal copies of documents, digital data. One method is to insert watermarked image
also to check the number of commercials in a Cable in spatial domain, the other is to insert in the frequency
Channel. Watermarks embedded into multimedia domain. Inserting watermark in the spatial domain has
for enforcing copyrights must uniquely identify the many disadvantages as compared to the frequency
data and must be difficult to remove, even after domain. One of the major drawbacks is that image is
various signal-processing tools. Now a days blurred while inserting watermark in spatial domain.
compression is an integral part of the signal This type of watermarking is also less robust to various
processing. In our paper we present a technique for signal processing techniques, for example compression,
embedding a watermark, which is robust in both filtering etc. In this paper we used two transformations
compression schemes. We have exploited the to pick a specific band of frequency coefficients that is
properties of Human Visual system (HVS) to select a more robust to any compression technique i-e DCT as
band of frequencies so that there is least visual well as Wavelet based compression technique.
difference between the original and the
watermarked image. Both the Discrete Cosine
2.1 Watermark Embedder
Transform and Discrete Wavelet Transform
technique have been incorporated to make sure that
First the whole image is transformed from spatial to
even after both types of compression, watermark is
frequency domain by Discrete Cosine Transform. DCT
still there in the image for copyright purposes. We
compacts the image energy in upper left corner of the
have optimize the algorithm so that it works fine in
transform matrix. The selection of particular
real time too.
transformed coefficient for embedding the watermark is
based on the Human Visual System (HVS). As HVS is
more sensitive to lower frequencies than to higher
1 INTRODUCTION frequencies. By inserting watermark at higher
frequencies gives watermarked images visually equal to
Digital watermarking is the enabling technology to original one (less blur and less distorted), but the
prove the ownership of copyrighted material, detect the compression of image will disturb the watermark badly.
originator of illegally made copies, monitor the usage So we select the one fourth coefficient of the transform
of the copyrighted multimedia data. Basically, at medium frequencies. This is done by the
watermarks, labels or codes embedded into multimedia preprocessor block in the watermarked embedder block
data for enforcing a copyright must uniquely identify
the data as property of the copyright holder, and must

9
 diagram as shown in fig 1. Discrete Wavelet transform
is used to split this medium frequency to sub-bands.
Input The low-low band is selected for inserting the
Image watermarked image.
The embedder operates in such a way that it generates a
DCT pseudo-number according to a key and insert these
pseudo-number at low-low band only when a pixel
value of Watermark is zero. This process repeats equal
Tansformed
Image

Pre
Procesing

Watermarked
1/4 TC of Tr_Image Image

DCT
DWT
Key

LL LH Tansformed
Rand Image
HL HH
Pre
Procesing
Pn-Seq logo(k)
+ == 0
Yes No
1/4 TC of Tr_Image

LL LH DWT

HL HH
LL LH

IWT
HL HH
Logo(K)=0

Yes
Post
Procesing
Correlation >Threshold

PN-Seq No
IDCT
Rand Logo(K)=1
Watermarked
Image
Key

“Figure 1 – Watermark Embedder” ═Figure 2 – Watermark Detector”

10
 to the size of the logo. After adding pseudo-number at
low-low band, the image is reconstructed ,i-e IDWT is
taken and then low and high frequencies transform
coefficient are concatenated .This is done by the post
processor component in the block diagram of embedder
in fig 1. After that Inverse Discrete Cosine Transform is
taken to form watermarked image.

2.2 Watermark Detector

Here we have proposed blind watermark detection in

which we require neither original image nor logo
image. We have only the key that was used in pseudo-
number generation. These pseudo-numbers are
correlated with a low-low frequency band and the result
is stored in an array as shown in fig 2. This is repeated
until logo size. A threshold value is selected to decide
whether the logo has pixel value of “1” or “0”.If
“Figure 3 Original Image
correlated value is greater than threshold ,it means that
there is a zero in the logo at that position otherwise it is
“1”.This process reconstruct the whole logo.

3 SIGNAL TRANSFORMATION
ATTACKS
This watermark has been tested for the following signal
transformation:
I. Watermarked image was encoded using JPEG
(Discrete Cosine Transform based
compression technique) and then decoded and
watermark was retrieved successfully from the
image. The original size of the watermarked
image was 258kb as shown in fig3. After
compression, its size was 42kb. The
compressed image was decoded and
watermark was still detectable.
II. Watermarked image was encoded using
Figure 4. Watermarked Image”
JPEG2000 (wavelet based compression
technique) and again after decoding,
watermark was retrieved successfully. Size of
compressed image was 36kb.
III. Gaussian Random Noise of zero mean and
variance “3” was added to the watermarked
image and after that, watermark was still
detected successfully.
IV. Low Pass filtering was applied to
watermarked image and after that, watermark
was still detectable.
V. High Pass Filtering was applied to
watermarked image and it proved to be robust ═Figure 5 – Digital Signature used as Watermark”
for this signal transformation process too.

11
4 CONCLUSIONS
In this paper, an algorithm is presented for embedding
of Watermark in an image and its extraction. In our
technique, we have selected a moderate band of
frequencies which are neither so low that it becomes
visible nor so high that it gets filtered out. Also it is
robust for classical as well as multi-resolution
compression techniques. The watermark has been tested
for certain attacks and it has proved to be robust enough
to use it in any of the compression techniques. At the
same time, the algorithm works fine in real time
environment and it takes the least processing power.

5 REFERENCES
[1] G. Voyatzis, N. Nikolaidis and I. Pitas, “Digital
watermarking: An overview,” EUSIPCO-98, Greece, Sept.
1998, vol. 1, pp. 9-12.
[2] R. G. van Schyndel, A.Z.Tirkel, N.R.A Mee, C.F.
Osborne, “A digital watermark,” Proceedings of the IEEE
International Conference on Image Processing, Austin, Texas,
USA, November, 1994, vol. 2, pp. 86- 90.
[3] R. G. van Schyndel, A. Z. Tirkel, C. Osborne, “Towards a
robust digital watermark,” Proceedings of the ACCV-95
Conference, Nanyang Technological University, Singapore,
December 5-8, 1995.
[4] A. Tirkel et al., “Electronic Watermark,” Digital Image
Computing, Technology and Applications—DICTA 93,
Maquarie University, Sidney, 1993, pp. 666-673.
[5] I. Cox et al., “Secure Spread Spectrum Watermarking for
Multimedia,” IEEE Trans. on Image Processing, Vol. 6, No.
12, Dec. 1997, pp. 1673-1687 (originally published in 1995 as
NEC TR 95-10).
[6] N. Kaewkamnerd and K.R.Rao, “Wavelet based
watermarked image detection using Multiresolution image
registration method,” TENCON 2000, Kuala Lumpur,
Malaysia, Sept. 2000.

12