See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/26511487

Digital Watermarking in Wavelet Transform Domain

Article  in  Radioengineering · January 2001

Source: DOAJ

CITATIONS READS

31 169

3 authors, including:

Emil Matus Dusan Levicky

Technische Universität Dresden Technical University of Kosice - Technicka univerzita v Kosiciach
49 PUBLICATIONS   393 CITATIONS    78 PUBLICATIONS   360 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

5G SDR architecture View project

All content following this page was uploaded by Dusan Levicky on 06 June 2014.

The user has requested enhancement of the downloaded file.

Radioengineering Digital Watermarking in Wavelet Transform Domain 1
Vol. 10, No. 2, July 2001 M. ČANDÍK , E. MATÚŠ, D.LEVICKÝ

DIGITAL WATERMARKING
IN WAVELET TRANSFORM DOMAIN
2. Digital Watermarking
Marek ČANDÍK, Emil MATÚŠ, Dušan LEVICKÝ
Dept. of Electronics and Multimedia Communication A watermark is hidden information within a digital
Technical University of Košice signal. For the watermarking several techniques have been
Park Komenskeho 13, 041 20 Kosice developed. Watermarking technique can be divided into
Slovak Republic two main groups:
• spatial domain watermarking,
• frequency domain watermarking.
Abstract Techniques that work in spatial domain can suffer
from signal compression and hostile attacks [2]. Frequency
This paper presents a technique for the digital domain techniques are much more robust against compres-
watermarking of still images based on the wavelet sion and geometrical transformations than spatial domain
transform. The watermark (binary image) is embedded techniques. Nevertheless, one weakness for may spatial fre-
into original image in its wavelet domain. The original quency approaches is that the human visual system is not
unmarked image is required for watermark extraction. taken into account when selecting positions to insert the
The method of embedding of digital watermarks in watermark. Because of the invisibility constraint of a wa-
wavelet transform domain was analysed and verified termark, these techniques have to use signals of relatively
on grey scale static images. lower power than would otherwise be possible, to avoid
degrading the image quality, inevitably limiting the robust-
ness of the watermark [2].
Proposed properties were shown that for watermarked
media several requirements must be satisfied [6],[7]:
Keywords • Imperceptibility – the watermark should be impercep-
Wavelet transform, digital watermarking tible, not to affect the viewing experience of the image
or the quality of signal.
• Undeleting – the watermark must be difficult or even
1. Introduction impossible to remove by a hacker, at least without ob-
viously degrading the host signal.
Digital representations of copyrighted material such • Statistically undetection – A pirate should not be able
as movies, songs, and photographs offer many advantages. to detect the watermark by comparing several water-
However, the fact that an unlimited number of perfect marked signals belonging to the same author.
copies can be illegally produced is a serious threat to the
rights of content owners. Until recently, the primary tool • Robustness – The watermark should be survive by the
available to help protect content owners’ rights has been using of the lossy compression techniques and signal
encryption. Encryption protects content during the processing operations (signal enhancement, geometric
transmission of the data from the sender to receiver. image operations, noise [8], filtering [9], etc.).
However, after receipt and subsequent decryption, the data Robustness is crucial to the success of watermark em-
is no longer protected and is in the clear. Watermarking bedding. To achieve an imperceptible watermarking is not
complements encryption. A digital watermark is a piece of difficult by minor modification of the host data. Making the
information that is hidden directly in the media content, in watermark indestructible, however, is not a trivial problem.
such a way that it is imperceptible to a human observer, but
easily detected by a computer. The principal advantage of The process of image watermarking can be represen-
this is that the content is inseparable from the water- ted by the addition of a noise term that is a function of the
mark [1]. watermark signal, w, and possibly of the original image, I.
Watermarked image, I’, can be created in wavelet trans-
They are some ways to watermark embedding into form domain. When an image undergoes wavelet decompo-
image information. One of the possible method is using of sition, its components are separated into bands of approxi-
discrete wavelet transform for watermark embedding into mately equal bandwidth on a logarithmic scale much as the
an image. In this paper is presented an algorithm of digital retina of the eye splits an image into several components. It
watermarking in wavelet domain. is, therefore, expected that use of discrete wavelet trans-
Radioengineering Digital Watermarking in Wavelet Transform Domain 2
Vol. 10, No. 2, July 2001 M. ČANDÍK , E. MATÚŠ, D.LEVICKÝ

form will allow the independent processing of the resulting orthogonal function, which can be applied to a finite group
components much like the human eye. of data [4]. Wavelets are functions that satisfy certain re-
quirements [5]
The algorithm of embedding of digital watermark in
frequency domain, in generally is shown on Fig. 1. • localized in time and frequency,
• integrate to zero,
Image Watermark
• quick and easy calculation of the direct and inverse
Transformation Pseudorandom Permutation wavelet transform.

Permuted
They are various wavelets: Haar, Coiflet, Daubechie,
Frequency
domain Watermark etc. Whereas the basis function of the Fourier transform is
a sinusoid, the dyadic wavelet basis is a set of function
which are defined by a recursive difference equation

+ M −1
φ ( x ) = ∑ ck φ (2 x − k ) , (1)
Inverse Transformation k =0

Watermarked
Image where M is the number of nonzero coefficients. The value
of coefficients is determined by constraints of
orthogonality and normalization.
Fig. 1 Process of digital watermarking in frequency domain
Wavelet transform uses wavelets as basis and is a tool
that cuts up data or functions or operation into different fre-
For watermark extraction is needed watermarked image quency components, and then studies each components
and also the original image. Than the process of watermark with a resolution matched to its scale [5].
extraction can be realized by following Fig. 2.
The discrete representation of fast wavelet transform
is used for image processing. DWT is defined by
Image Watermarked Image
j

∑ f (i )φ (2 )
−
F ( j, k ) = 2
transformation transformation −j
2
i−k (2)
Frequency domain Frequency domain ∀i

inverse DWT is defined by

- j

Watermark
Inverse Image Permutation
−
f (i ) = ∑∑ F ( j , k ) 2 2 φ 2 − j i − k ( ) (3)
∀j ∀k

Fig. 2 Process of watermark extraction

where i, j, and k are integers.
The 2D algorithm is based on separable variables lea-
ding to prioritising of x and y directions, and scaling func-
tion is defined by
3. Wavelet transform
φ (x, y ) = φ (x )φ ( y ) . (4)
Wavelets are functions that satisfy certain mathema-
tical requirements and are used in representing data or ot- The wavelet decomposition of original image I is
her functions. The idea is not new. Approximation using shown on the Fig. 3.
superposition of functions has existed since early 1800’s ,
when Joseph Fourier discovered that he could superpose The original image is decomposed by the lowpass
sines and cosines to represent other functions. However, in (LP) and highpass (HP) filters followed by downsampling
wavelet analysis, the scale that we use to look at data plays first of rows and then of columns. The result of wavelet
a special role. Wavelet algorithms process data at different decomposition is approximation of original image and
scales and resolutions. If we look at a signal with a large three detail signals (horizontal, vertical and diagonal).
“window”, we would notice gross features. Similarly, if we Process of wavelet decomposition can be expressed in the
look at a signal with a small ”window”, we would notice following form
small features. The result in wavelet analysis is to see both cI j = cI j +1 + cD (jh+1) + cD (jv+)1 + cD (jd+1) (5)
the forest and the trees.
The wavelet transform is a linear transform that gene- The next step is decomposition of cIj+1 and so on.
ralizes the properties of the Haar transform. A wavelet, in After j decomposition steps we obtain decomposed image
the sense of the Discrete Wavelet Transform (DWT), is an in the next form (j=2, see Fig. 4).
Radioengineering Digital Watermarking in Wavelet Transform Domain 3
Vol. 10, No. 2, July 2001 M. ČANDÍK , E. MATÚŠ, D.LEVICKÝ

4. Wavelet based Watermarking

Watermarking in the DWT domain consists of enco-
ding and decoding parts.
In the encoding part, we first decompose an image
into wavelet frequency domain and we obtain decomposed
image. Image permuted watermark we add to obtained ima-
ge decomposition. The watermark permutation is reversible
and it is key for correct watermark extraction. For each
coefficient within the wavelet domain, the key has a corres-
ponding value of one or zero (if watermark is an binary
image) to indicate if the coefficient is to modify or not. No-
te that watermarks are not inserted into the LH1, HL1 and
HH1 bands (where L denotes the lowpass band and H
Fig. 3 Wavelet decomposition of image denotes the highpass band), because the energies in these
bands are relatively small.
In decoding part, we then take the two-dimensional
(2D) inverse DWT (IDWT), obtaining the watermarked
image I’.

5. Experiments
Our experiments we have realized on static gray-scale
image Einstein in *.pgm format. As a watermarks were
employed binary images that are shown in following figu-
res (Fig. 6, Fig. 7).

32 x 32

Watermark1 Watermark2

Fig. 6 Employed watermarks

Fig. 4 Example of wavelet decomposed image

64 x 64
The process of image reconstruction is shown in Fig. 5.

Watermark4 Watermark3

Fig. 7 Employed watermarks

In our experiments, the two-level DWT is employed

for watermarks with size 64x64, in which an image is
decomposed into 7 subbands, and for watermarks with size
32x32 was used three-level DWT, in which an image is
decomposed into 10 subbands.
Quality of watermarked image was analyzed with
objective and subjective criteria of image quality. As objec-
tive criteria of image quality was used Peak –Signal –to –
Noise Ratio defined as following relation (6)
Fig. 5 Image reconstruction
 Radioengineering Digital Watermarking in Wavelet Transform Domain 4
Vol. 10, No. 2, July 2001 M. ČANDÍK , E. MATÚŠ, D.LEVICKÝ

255 2 (6)
PSNR = 10 log10
⎛ 1 ⎞ N1 N 2
References
⎜⎜ ⎟⎟∑∑ (I (i, j ) − I ' (i, j ))2
⎝ N 1 .N 2 ⎠ i =1 j =1 [1] MILLER, M. L., COX, I. J., LINNARTZ, J. P., KALKER, T.: A
review of watermarking principles and practices, 1999,
http://www.neci.nj.nec.com/
Experimental results are shown in following Tab. 1,
and an example of watermarked image is shown on Fig. 8. [2] TAO, B., DICKINSON, B.: Adaptive watermarking in the DCT
domain., IEEE Int. Conf. ASSP ’97, 1997.

PSNR [dB] [3] COX, I. J., MILLER, M. L.: A review of watermarking and the
importance of perceptual modelling. Proc. of Electronic Ima-
Watermark1 (32 x 32) 60,33 ging ’97, 1997.
Watermark2 (32 x 32) 55,97 [4] EDWARDS, T.: Discrete wavelet transforms: Theory and im-
Watermark3 (64 x 64) 50,89 plementation. http://sinh.stanford.edu/

Watermark4 (64 x 64) 50,36 [5] JUN, J.: Introduction to wavelet transform., 1999,
http://ic.hansung.ac.kr/

Tab. 1 Experimental results [6] LUKÁČ, R.: An Adaptive Control of LUM Smoother. Radio-
engineering, vol. 9, no. 1, April 2000, pp.9-12.

[7] LUKÁČ, R. - MARCHEVSKÝ, S.: Adaptive LUM Smoother

Controlled by Adaptive Threshold System. Journal of Electrical
Engineering, no. 3-4, vol. 51, 2000, pp.100-104.

About authors...
Marek ČANDÍK was born in Košice, Slovakia, 1974. He
graduated from the Technical University of Košice in 1997,
then he started Ph.D. study at Department of Electronics
and Multimedia Communications, Faculty of Electrical
Engineering and Informatics, Technical University of Ko-
šice. His work includes digital image processing and trans-
mission.

Emil MATÚŠ was born in Gelnica, Slovakia, 1965. He

received the M.Sc. and PhD. degrees in electrical engi-
Fig. 8 Example of watermarked image neering at the Technical University of Košice. Currently he
is the assistant professor in Department of Electronics and
Multimedia Communications, Faculty of Electrical Engi-
neering and Informatics, Technical University of Košice.
6. Conclusions His research interests are image coding, image transmis-
Digital watermarking of still images based on the sion, wavelet transform.
wavelet transform is a good method for authentication of Dušan LEVICKÝ was born in Slanec (Slovak Republic)
image materials. The using of wavelet transform satisfies in 1948. He received the M.Sc. and PhD. degrees at Tech-
requirements for watermarked media. We demonstrate the nical University of Košice and now he is proffesor at the
investigation of image distortion on the relation between Departement of Electronics and Multimedia Communica-
the original image and watermarked image. In proposed tions Technical University in Košice. His research interests
algorithm the original unmarked image is required for include digital image processing, image transmission, and
watermark extraction. cryptography.

View publication stats