WAVELET BASED AUTHENTICATION/SECRET TRANSMISSION THROUGH IMAGE RESIZING (WASTIR)

Signal & Image Processing : An International Journal (SIPIJ) Vol.4, No.2, April 2013
DOI : 10.5121/sipij.2013.4205 59
WAVELET BASED AUTHENTICATION/SECRET
TRANSMISSION THROUGH IMAGE RESIZING
(WASTIR)
Madhumita Sengupta1
and J. K. Mandal2
1, 2
Department of Computer Science and Engineering, University of Kalyani
Kalyani, Nadia, Pin. 741235, West Bengal, India
1
madhumita.sngpt@gmail.com 2
jkm.cse@gmail.com
ABSTRACT
The paper is aimed for a wavelet based steganographic/watermarking technique in frequency domain
termed as WASTIR for secret message/image transmission or image authentication. Number system
conversion of the secret image by changing radix form decimal to quaternary is the pre-processing of the
technique. Cover image scaling through inverse discrete wavelet transformation with false Horizontal and
vertical coefficients are embedded with quaternary digits through hash function and a secret key.
Experimental results are computed and compared with the existing steganographic techniques like WTSIC,
Yuancheng Li’s Method and Region-Based in terms of Mean Square Error (MSE), Peak Signal to Noise
Ratio (PSNR) and Image Fidelity (IF) which show better performances in WASTIR.
KEYWORDS
WASTIR; Authentication; Steganography; Watermarking; DWT; MSE; PSNR; IF.
1. INTRODUCTION
Digital steganography plays a vital role in verifying the integrity of digital information.
Watermarking can achieve copyright protection, ownership verification or secret message
transmission depending on the ways of data hiding. Various techniques are available categorized
under spatial domain [2, 5, 13], frequency domain [3, 5] or composite domain [4]. Like LSB,
PDV, SVD [1], DCT [4], DFT, DWT [1, 3, 5, 12], techniques of arbitrary resizing [12] and many
more.
Necessity of security is a major concern of today’s researchers, despite of various works in this
area still there is dearth of investigation in the subject. There are two major improvements seen in
the proposed one. Firstly the degradation of the quality of the cover image is less compared to
other techniques. Secondly the payload has been increased considerably without degrading the
fidelity of the image.
Various parametric tests have been performed and results obtained are compared with most recent
existing techniques such as, WTSIC [3], Yuancheng Li’s Method [12] and Region-Based
watermarking [13], based on Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and
Image Fidelity (IF) analysis [6] which shows a consistent relationship with the existing
techniques.
Section 2 of the paper deals with the proposed scheme. Resizing through discrete wavelet
transformation technique with embedding has been described in section 3, Results and

60
discussions are outlined in section 4, conclusions are drawn in section 5 and that of references are
given at end.
2. THE SCHEME
Digital images are represented as a two dimensional matrix of pixel intensity values in the form
of integer representation in the range of 0 to 255. The proposed WASTIR technique is divided
into two major groups of activity shown in figure 1 and 2.
The schematic representation of activity at sender is shown in figure 1. The cover image of
dimension N x N passes through the process of inverse DWT to generate translated image with
predefined horizontal, vertical and diagonal information of N x N dimension each, as describe in
section 3.3. The secret message/image converts its pixel intensity values from decimal to
quaternary with radix 4, as shown in figure 4.
On applying 2 x 2 masks over translated image R00, R01, R10 and R11 values are fetched to embed
two digits (4 bits) of quaternary system based on hash function as explained in section 3.4.
Fidelity adjustment has been incorporated to preserve the visual quality of cover image which is
discussed in section 3.5 to generate stego-image.
Figure 1. Embedding Algorithm

61
During decoding the stego-image passes through the extraction algorithm based on same hash
function to regenerate the secret message/image. On authentication or secret message generation
the stego-image passes through the forward DWT as shown in section 3.1. The output of forward
DWT generates the average coefficients of N x N dimension that is exactly the cover image taken
by sender with MSE zero and PSNR infinity. Embedding process is shown in figure 2.
Figure 2. Extraction/Authentication Algorithm

62
3. RESIZING THROUGH TRANSFORMATION AND EMBEDDING
In image processing each transform equation is available as pair which is reversible and termed as
forward and inverse transformation respectively[7]. In Wavelet based forward transformation the
image converts from spatial domain to frequency domain using (1) and (2), and in inverse
transformation the reverse procedure is followed (3). Mathematically the image matrixes multiply
with scaling function coefficients and wavelet function coefficients in the formation of
transformation matrix [8].
Where x[n] is original signal, h[x] is half band low pass filter, g[x] is Half band high pass filter,
YLow[k] is output of high pass filter after sub sampling by 2, YHigh[k] is output of low pass
filter after sub sampling by 2.
3.1 Forward Transformation
In the proposed technique Mallat based two-dimensional wavelet transform is used in order to
obtain set of bi-orthogonal subclasses of images [9]. In two-dimensional wavelet transformation,
a scaling function φ(x, y) is used represent by (4).
φ (x, y) = φ(x) φ(y)
and if ψ(x) is a one-dimensional wavelet function associated with the one–dimensional scaling
function φ(x), three two dimensional wavelets may be defined as given in (5). Figure 3 represents
functions in visual form.
ψH (x,y)= φ (x) ψ(y)
ψV (x,y)= ψ (x) φ(y)
ψD (x,y)= ψ (x) ψ(y)
Low resolution
sub-image
average ‘A’
ψ (x, y)= φ(x) φ(y)
Horizontal
Orientation sub-
image ‘H’
ψH
(x,y)= φ(x) ψ(y)
Vertical
Orientation sub-image
‘V’
ψV
(x, y)= ψ(x)φ(y)
Diagonal
Orientation sub-
image ‘D’
ψD
(x,y)=ψ (x) ψ(y)
Figure 3. Image decomposition in Wavelet transforms
(1)
(2)
(3)
(4)
(5)

63
As per Haar forward transform scaling function coefficients and wavelet function coefficients [8]
H0 = ½, H1 = ½, G0 = ½ G1 = -½ are taken. In WASTIR stego-image at receiver side of
dimension 2N x 2N on forward DWT breaks into four sub images as shown in figure 3. Each sub-
image is of dimension N x N. Out of which low resolution/average sub image on comparison with
is the original cover image taken by sender, generates MSE zero and that of PSNR infinity. Thus
without hampering the original image, secret message/image can be transmitted through
unsecured network, securely.
3.2 Inverse Transformation
Inverse transformation is just the reverse of the forward transformation with column
transformation done first followed by row transformation. But the coefficient values are different
for column/row transformation matrices. The coefficient for reverse transformation are H0 = 1, H1
= 1, G0 = 1, G1 = -1 [8]. In the first phases of proposed WASTIR technique, original image of
dimension N x N passes through inverse transform to generate translated image matrix of
dimension 2N x 2N.
For any inverse wavelet transformation (equation (3) of section III) four equal sized coefficient
matrix is required such as low resolution sub-image ‘A’, horizontal orientation sub-image ‘H’,
vertical orientation sub-image ‘V’ and diagonal orientation sub-image ‘D’ as shown in figure 3,
they are output of forward wavelet transformation. Thus to perform inverse wavelet
transformation without forward wavelet transformation three false coefficient matrix ‘H’, ‘V’,
and ‘D’ of same size is required with original image as ‘A’ for the process of resizing or scaling.
3.3 Resizing/Scaling
To scale image by a factor T, up-sampling by a factor of T is performed. This up-sampling is
done through discrete wavelet transformation, where four types of coefficient sets are voted
through (3) to generate image of dimension TN x TN. The original cover image of dimension N x
N is taken as average ‘A’, where as horizontal, vertical and diagonal coefficients are taken as
predefined values. The space overhead will depend on the value of T.
Let us take an example where resize of the image has been done by a factor of 2, for which cover
image of dimension 2 x 2 is taken and substitute with 2 x 2 dimension of horizontal, vertical and
diagonal sub components each with a false value β. Then after inverse wavelet transformation,
translated image generates with a dimension of 4 x 4 as shown in figure 4.
X Y
H = β
X0 X1 Y0 Y1
Z W X2 X3 Y2 Y3
V = β D = β
Z0 Z1 W0 W1
Z2 Z3 W2 W3
Figure 4. Resizing of 2 x 2 cover image into 4 x 4 translated image through inverse wavelet
transformation
Where β is a user define integer taken as a false horizontal, vertical and diagonal coefficients
to perform IDWT.
3.4 Embedding
The secret image of dimension 512 x 256 is selected as shown in figure 8.k. This secret image is
of 1048576 bits in size. Each pixel intensity value represented in the form of decimal number
system in range of 0 to 255 has been converted to quaternary number system. The minimum
I DWT

64
intensity value 0 in decimal can be represented as (0000)4 in quaternary number system, and 255
the maximum intensity supported by our system of image can be represented as (3333)4. Few
examples with conversion technique are given in figure 5. After conversion of radix the intensity
values of secret image is stored in a long chain of digits array ranging from 0 to 3.
(255)10 Radix 10(Decimal number system)
(3333)4 Radix 4(Quaternary number system)
(159)10 = 2 * 43
+ 1 * 42
+ 3 * 41
+ 3 * 40
= (2133)4
(146)10 = 2 * 43
+ 1 * 42
+ 0 * 41
+ 2 * 40
= (2102)4
(3333)4 = 3 * 43
+ 3 * 42
+ 3 * 41
+ 3 * 40
= (255)10
(0301)4 = 0 * 43
+ 3 * 42
+ 0 * 41
+ 1 * 40
= (49)10
Figure 5. Example of conversion of number system
A 2 x 2 mask on translated image is used which provides four values at every pair of coordinates,
labeled as R00, R01, R10, and R11. Out of four two diagonal coordinate pairs, R01 and R10 are
selected for embedding.
The hash function with mod operation works together to select the embedding position and the
embedded value, as per (6). This procedure is illustrated in figure 6. The hash function selects R01
as the first embedding position. Then ‘R01 mod 4’ operation is performed and the difference is
calculated between the secret quaternary digit and mod result. If the difference is positive and
more than zero then the new value for the R01 position is recalculated by subtracting the
difference. Else if the difference is less than zero then the new value for the position is
recalculated by adding the difference.
diff =‘Secret digit’ – (‘old value of embedding position’ % 4)
If (diff > 0 ) then
‘New value of embedding position’= ‘Old value of
embedding position’– (4 – diff)
if (diff < 0 ) then
‘New value of embedding position’= ‘Old value of
embedding position’+ (4 + diff)
Hence two secret digits 3 and 2 taken from the chain of secret digits are embedded through hash
function into R01 and R10 based on generalized (6).
Let’s take an example of secret quaternary
digit chain: 3 2 3 0 2 3………
236 236 236 235
236 236 234 236
2 x 2
translated
image values
Embedded
with two
secret digits
if [(‘secret digit’– (R01 % 4)) > 0] then
[R01= R01 – (4 – (secret digit’– (R01 % 4))
else if [(‘secret digit’– (R01 % 4)) < 0] then
[R01= R01 + (4 + (secret digit’– (R01 % 4))
Figure 6. Embedding two secret digits in 2 x 2 mask of translated image.
(6)

65
3.5 Adjustment
Diminutive adjustment is required to preserve the near original image intensity. The adjustment
enforces equal amount of deduction from the diagonal value R11. The embedded image peruses
forward DWT without adjustment as shown in figure 7.c where the low resolution or average
value changes from 236 to 235.25 and with adjustment as shown in figure 7.d, where the low
resolution or average value retains at 236.
236 235 236 235
234 236 234 239
235.25 -0.25
0.25 -0.75
236 -1
-0.5 1.5
c. Cover image after
forward DWT Without
adjustment
d. Cover image after
forward DWT With
adjustment
Figure 7. Procedure of diminutive adjustment
From figure 7 it is clear that, to regenerate the original cover image the low resolution sub-image
must be intact on forward DWT, after embedding also, which generates the requirement of
adjustment.
Adjustment is a simple calculation of equating the total average before embedding and after
embedding by readjusting the value of R11. In the above example of figure 6 and figure 7, the
average of 2 x 2 translated image is 236. After embedding two secret digits the new average will
become 235.25. Thus adjustment needs to add value 3 to R11 to equate the average value and
bring it back to 236.
3.6 Authentication
At the receiving end the setgo-image passed through 2 x 2 mask labeled as R00, R01, R10 and R11.
Mod operation on R01, and R10 by 4 to fetch the secret quaternary digits using the function given
in (7).
Secret digit1 = R01 % 4
Secret digit2 = R10 % 4
After extraction of all the quaternary digits, the numbers are converted to radix 10 and the secret
message/image is generated which when compared with the original secret image will
authenticate the cover image.
4. RESULTS AND DISCUSSIONS
Ten PPM [10] images have been taken and applied on WASTIR to obtain results. All cover
images are 512 x 512 in dimension and secret coin (k) used as authenticating image of 512 x 256
in dimension. The images are shown in figure 8.
a. Embedded with two
secret digits without
adjustment
b. Embedded with two
secret digits with adjustment
(7)

66
a. Baboon b. Boat c. Clock d. Couple
e. Elaine f. Jet g. Map h. Space
i. Tank j. Truck k. Secret Coin Image
Figure 8. Cover Images of dimension 512 x 512 and Secret Image 512 x 256
Table 1 shows the MSE, PSNR and IF for ten benchmark gray scaled images for embedding two
quaternary digits into a 2 x 2 masks of translated image. As every digit requires maximum two
bits to represent thus in total 1.0 bpB (bit per byte) of payload is supported by the proposed
WASTIR algorithm. From table 1 it is clear that average MSE over ten images are 4.6478536,
that of PSNR is 41.4588557 and IF is 0.999762.
Table I. Statistical data on applying WASTIR over 10 gray scale translated images of 1024 x
1024 dimensions with secret coin as secret message/image.
Cover Image
512 x 512
MSE PSNR IF
(8.a) Baboon 4.615559 41.488561 0.999752
(8.b) Boat 4.643509 41.462341 0.999756
(8.c) Clock 4.714600 41.396355 0.999875
(8.d) Couple 4.572992 41.528799 0.999727
(8.e) Elaine 4.631128 41.473935 0.999777
(8.f) Jet 4.796215 41.321817 0.999847
(8.g) Map 4.623446 41.481146 0.999865
(8.h) Space 4.697512 41.412125 0.999724
(8.i) Tank 4.506279 41.592623 0.999753
(8.j) Truck 4.677296 41.430855 0.999616
A comparative study has been made between WTSIC [3] and WASTIR in terms of Mean
Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and bits per Byte as in table 2. On
increasing the bits per byte from 0.5 to 1.0, exactly double, the MSE increases by 0.599707 and
PSNR increases by 0.61696 dB only which is acceptable compared to bits per Byte ratio.

67
Table II. Comparison between WTSIC, Horizontal sub-image method with proposed
WASTIR.
Image
WTSIC, Horizontal
Orientation sub-image
WASTIR
0.5 (bit per Byte) 1.0 (bit per Byte)
MSE PSNR MSE PSNR
Elaine 3.786247 42.348714 4.631128 41.473935
Boat 3.696224 42.453221 4.643509 41.462341
Clock 4.426640 41.672001 4.714600 41.396355
Map 3.778519 42.357588 4.623446 41.481146
Jet 4.722733 41.388870 4.796215 41.321817
Comparison of the proposed technique has also been made between Yuancheng Li’s Method [12],
Region-Based watermarking methods [13] and ATFDD [14] as shown in table III. In all of the
cases optimized results are obtained in proposed methods in terms of PSNR and payloads
compare to existing techniques. But in case of comparison with WTSIC [3], on embedding into
large images with increase of payload by more than eight times, the PSNR has decreased by only
0.5 dB.
Table III. Comparison between various techniques with proposed WASTIR
Technique
Hiding
Capacity
(bytes)
Size of cover
image
bpB (Bits
per bytes)
PSNR
(dB)
Li’s Method[12] 1089 257*257 0.13 28.68
Region-Based[13] 16384 512*512 0.5 40.79
WTSIC[3] 16384 512*512 0.5 42.04
ATFDD[14] 32768 512 * 512 1.0 36.70
WASTIR 131072 1024*1024 1.0 41.43
5. CONCLUSIONS
In this paper we have addressed the issue of hiding data in cover image in frequency domain,
through discreet wavelet transformation technique (DWT). This method firstly resizes the image
arbitrarily by up-sampling, followed by embedding. The original cover image is regenerates at
the receiver by down sampling the image through transformation. The proposed technique
provides a good fidelity in terms of PSNR, IF compared to existing techniques.
ACKNOWLEDGEMENTS
The authors express deep sense of gratuity towards the Dept of CSE University of Kalyani where
the computational resources are used for the work and the PURSE scheme of DST, Govt. of
India.
REFERENCES
[1] S.Ramakrishnan, T.Gopalakrishnan, K.Balasamy “A Wavelet Based Hybrid SVD Algorithm For
Digital Image Watermarking”. Signal & Image Processing, (SIPIJ) Vol.2, No.3, DOI:
10.5121/sipij.2011.2313, pp 157- 174, September 2011.
[2] J. K. Mandal, M. Sengupta, “Steganographic Technique Based on Minimum Deviation of Fidelity
(STMDF)”, 2nd International Conference on EAIT, IEEE, doi:10.1109/EAIT.2011.24. pp. 298-301,
2011.

68
[3] J. K. Mandal, M. Sengupta. “Authentication / Secret Message Transformation through Wavelet
Transform based Subband Image Coding (WTSIC)”, IEEE, International Symposium on Electronic
System Design, Bhubaneswar, India. Print ISBN: 978-1-4244-8979-4. pp. 225-229.
doi:10.1109/ISED.2010.50, 2010.
[4] M. Sengupta, J. K. Mandal, “Self Authentication of Color Images through Discrete Cosine
Transformation (SADCT)”, IEEE catalog no : CFP1122P-CDR, ICRTIT, pp. 832-836, Anna
University, Chennai, ISBN No-: 978-1-4577-0589-2, 3rd-5th June 2011.
[5] M. Sengupta, J.K. Mandal, “Transformed IRIS Signature fabricated Authentication in Wavelet based
Frequency Domain (TISAWFD)”, International Journal of Advanced Research in Computer Science,
2 (5), pp- 486-490, ISSN No. 0976-5697, Sept –Oct, 2011.
[6] M. Kutter, F. A. P. Petitcolas, A fair benchmark for image watermarking systems, Electronic Imaging
'99. Security and Watermarking of Multimedia Contents, vol. 3657, Sans Jose, CA, USA, January
1999. The International Society for Optical Engineering,
http://www.petitcolas.net/fabien/publications/ei99-benchmark.pdf. (Last accessed on 25th March,
2011).
[7] The Wavelet Tutorial by Robi Polikar, http://users.rowan.edu/~polikar/WAVELETS/ WTpart1.html,
“Fundamental Concepts & an Overview of the Wavelet Theory”. The Wavelet Tutorial is hosted by
Rowan University, College of Engineering Web Servers, (Last accessed on 25thMay, 2011).
[8] Ian Kaplan, http://www.bearcave.com/misl/misltech/ wavelets/matrix /index.html, January 2002 (Last
accessed on 25th May, 2011.
[9] Marc Antonini, Michel Barlaud, Pierre Mathieu, and Ingrid Daubechies, “Image Coding Using
Wavelet Transform”, IEEE Transactions On Image Processing, vol. I, no. 2, April 1992.
[10] Allan G. Weber, The USC-SIPI Image Database: Version 5, Original release: October 1997, Signal
and Image Processing Institute, University of Southern California, Department of Electrical
Engineering. http://sipi.usc.edu/database/ (Last accessed on 25th May, 2012).
[11] E.-L. Tan, W.-S.Gan, S.K. Mitra, “Fast arbitary resizing of images in the discrete cosine transform
domain. IET Image Processing, vol.5 iss. 1, pp. 73-86, doi:10.1049/iet-ipr.2008.0004, 2011.
[12] Li Yuancheng, Xiaolei Wang, “A watermarking method combined with Radon transformand 2D-
wavelet transform”, IEEE, Proceedings of the 7th World Congress on Intelligent Control and
Automation, June 25 - 27, Chongqing, China, 2008.
[13] Nikolaidis, I. Pitas, “Region-Based Image Watermarking”, IEEE Transactions on Image Processing,
Vol. 10, no. 11, pp. 1721-1740, November 2001.
[14] M. Sengupta, J.K. Mandal, “An Authentication Technique in Frequency Domain through Daubechies
Transformation (ATFDD)”, International Journal of Advanced Research in Computer Science, 3 (4),
ISSN No. 0976-5697, pp - 236-242, July–August, 2012.
Authors
Madhumita Sengupta
M. Tech (CSE, University of Kalyani, 2010), pursuing Ph.D. as University Research
Scholar (University of Kalyani), in the field of transform domain based image
processing, steganography. Total number of publications is 14.
Prof. Jyotsna Kumar Mandal
M. Tech.(Computer Science, University of Calcutta), Ph.D.(Engg., Jadavpur
University) in the field of Data Compression and Error Correction Techniques,
Professor in Computer Science and Engineering, University of Kalyani, India. Life
Member of Computer Society of India since 1992 and life member of cryptology
Research Society of India. Ex-Dean Faculty of Engineering, Technology &
Management, working in the field of Network Security, Steganography, Remote
Sensing & GIS Application, Image Processing. 26 years of teaching and research
experiences. Nine Scholars awarded Ph.D. one submitted and eight are pursuing. Total
number of publications is more than two hundred seventy seven in addition of
publication of five books from LAP Lambert, Germany.

WAVELET BASED AUTHENTICATION/SECRET TRANSMISSION THROUGH IMAGE RESIZING (WASTIR)

More Related Content

What's hot (16)

Viewers also liked (19)

Similar to WAVELET BASED AUTHENTICATION/SECRET TRANSMISSION THROUGH IMAGE RESIZING (WASTIR) (20)

Recently uploaded (20)

WAVELET BASED AUTHENTICATION/SECRET TRANSMISSION THROUGH IMAGE RESIZING (WASTIR)