2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)
Color Image Steganography Technique Using
Daubechies Discrete Wavelet Transform
Ajaya Shrestha
Department of Electronics and Computer
ACEM
Kupondole-2, Lalitpur, Nepal
ajaya.shrestha@gmail.com
Abstract— The world in modern days is being more and more tied
up to the use of technology for facilitating day to day tasks. In this
regard, information security is turning to be great challenge when
sending information from one place to another with the aid of
technology. Steganography is one of the techniques for the safe
transmission which involves hiding information generally with other
information that only the receiver will know. In this research,
steganography technique using Daubechies Discrete Wavelet
Transform (DWT) is implemented. First the cover image is
transformed using Daubechies DWT and secret information is
embedded in coefficients of Daubechies DWT which gives stego
image. Reverse process is applied to obtain secret information from
stego image. The performance of the proposed approach is evaluated
using PSNR and MSE. Also analysis is done to find the best sub-band
to hide information among different sub-bands of Daubechies DWT.
Keywords—Stego Image; Discrete Wavelet Transform (DWT);
Peak Signal to Noise Ratio (PSNR); Mean Square Error (MSE)
I. INTRODUCTION
Transmission of information from one place to another always
has potential threats of being leaked before it reaches the
destination. Especially when one has to transmit secure and
confidential message, this risk is always high. To address
these threats, people always seek and invent technologies for
securely transmitting messages. One of the techniques is
information hiding.
There are many techniques of information hiding including
cryptography, watermarking and steganography.
Cryptography is art of protecting information by transforming
or encrypting it into an unreadable format, called cipher text.
Only those who possess a secret key can decipher or decrypt
the message into plain. Watermarking is a pattern
of bits inserted into a digital image, audio or video file that
identifies the file's copyright information (author, rights, etc.)
Steganography is an art of covert communication in which a
secret message is communicated by hiding it in a cover file, so
that the very existence of the secret message is not detectable.
The cover file can be image, audio or video; the most
commonly being the image files.
Steganography dates back to ancient Greece, where common
practices consisted of etching messages in wooden tablets and
978-1-4673-6744-8/15/$31.00@2015 IEEE
Dr. Arun Timalsina
Department of Electronics and Computer
IOE, Central Campus
Pulchowk, Lalitpur, Nepal
t.arun@ioe.edu.np
covering them with wax, and tattooing a shaved messenger's
head, letting his hair grow back, and then shaving it again
when he arrived at his contact point.
The advantage of steganography over cryptography alone is
that the intended secret message does not attract attention to
itself as an object of scrutiny. Plainly visible encrypted
messages no matter how unbreakable will arouse interest, and
may in themselves be incriminating in countries
where encryption is illegal. Thus, whereas cryptography is the
practice of protecting the contents of a message alone,
steganography is concerned with concealing the fact that a
secret message is being sent, as well as concealing the contents
of the message.
Secret Image
Embedding Extraction
Stego Image
Function Function
Cover Image Secret Image
Fig. 1. Basic Block Diagram of Steganography
II. RELATED THEORY
A. Discrete Wavelet Transform
The wavelet domain is growing up very quickly. Wavelets
have been utilized as a powerful tool in many diverse fields,
including approximation theory; signal processing, physics,
astronomy, and image processing. A wavelet is simply, a
small wave which has its energy concentrated in time to give a
tool for the analysis of transient, non-stationary or time-
varying phenomena. A signal can be better analyzed if
expressed as a linear decomposition of sums of products of
coefficient and functions. A two-parameter system is
constructed such that one has a double sum and coefficient
with two indices. The set of coefficients are called the DWT
of a signal.
DWT splits the signal into high and low frequency parts. The
high frequency part contains information about the edge
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)
components, while the low frequency part is split again into
high and low frequency parts which is shown in figure 2 [1].
Fig. 2. Block Diagram of 1-Step 2-D DWT
The high frequency components are usually used for
steganography since the human eye is less sensitive to changes
in edges [2]. In two dimensional applications, for each level of
decompositions, we first perform the DWT in the vertical
direction, followed by the DWT in the horizontal direction. As
we can see in figure 3 [2], after the first level of
decomposition, there are four sub-bands: LL, LH, HL and HH.
Fig. 3. First Level Wavelet Decomposition
B. Haar Wavelet
Haar wavelet is one of the oldest and simplest wavelet. It is
also the symmetric wavelet. In discrete form Haar wavelets
are related to a mathematical operation called the Haar
transform. The Haar transform works as a prototype for all
other wavelet transforms. In mathematics, the Haar wavelet is
a sequence of rescaled "square-shaped" functions which
together form a wavelet family or basis. Wavelet analysis is
homogeneous to Fourier analysis in that it allows a target
function over an interval to be represented in terms of an
orthonormal function basis. It represents the same wavelet as
Daubechies db1. The Haar wavelet transform has a number of
advantages such as it is conceptually fast, simple, memory
efficient, since it can be calculated in place without a
temporary array.
The Haar wavelet also has limitations. In generating each of
averages for the next level and each set of coefficients, the
Haar transform performs an average and difference on a pair
of values. Then the algorithm shifts over by two values and
calculates another average and difference on the next pair. The
high frequency coefficient spectrum should reflect all high
frequency changes. The Haar window is only two elements
wide. If a big change takes place from an even value to an odd
value, the change will not be reflected in the high frequency
coefficients.
978-1-4673-6744-8/15/$31.00@2015 IEEE
C. Daubechies Wavelet
Daubechies wavelets are the most popular wavelets. They
represent the foundations of wavelet signal processing and are
used in various applications. These are also called Maxflat
wavelets as their frequency responses have maximum flatness
at frequencies 0 and π. The Daubechies wavelet transforms are
defined in the same way as the Haar wavelet transform—by
computing running averages and differences via scalar
products with scaling signals and wavelets—the only
difference between them consists in how these scaling signals
and wavelets are defined. For the Daubechies wavelet
transforms, the scaling signals and wavelets have slightly
longer supports, i.e., they produce averages and differences
using just a few more values from the signal. This slight
change, however, provides a tremendous improvement in the
capabilities of these new transforms. The names of the
Daubechies family wavelets are written dbN, where N is the
order, and db the "surname" of the wavelet.
D. Peak Signal to Noise Ratio(PSNR)
It is the measure of reconstruction of the transformed
image. This metric is used for discriminating between
the cover and stego image which is given by equation 1
[13].
10 255 / (1)
E. Mean Square Error(MSE)
It is one of the most frequently used quality measurement
technique followed by PSNR. The MSE can be defined as the
measure of average of the squares of the difference between
the intensities of the stego image and the cover image. It is
popularly used because of the mathematical tractability it
offers. It is represented in equation 2 [13].
∑ ∑ ′
, , (2)
III. LITERATURE REVIEW
There are many steganography techniques which are capable
of hiding data within an image. These techniques can be
classified into two categories based on their algorithms: (1)
spatial domain based techniques; (2) transform domain based
techniques. The spatial domain based steganography technique
use either the LSB or Bit Plane Complexity Segmentation
algorithm. The most widely used technique to hide data is the
usage of the LSB [5]. The existing techniques are mainly
based on LSB (Least Significant Bit) where LSBs of the cover
file are directly changed with message bits. S. M Karim et al.
[6] has proposed a LSB technique for RGB true color image
by enhancing the existing LSB substitution techniques to
improve the security level of hidden information. LSB
matching image steganography and edge adaptive scheme was
proposed which can select the embedding regions according to
the size of secret message and the difference between two
consecutive pixels in the cover image. In [7] designing of
robust and secure image steganography based on LSB
insertion and RSA encryption technique has been used.
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)
M. A. Ahmed et al. [8] proposed a method in which a message
hidden inside an image by using the Least Significant Bit
technique and after creation of the hidden message, the image
will pass it in hash function to obtain hashing value using the
MD5 technique. In [9] a hash based approach proposed for
secure keyless steganography in lossless RGB images that an
improved steganography approach for hiding text messages in
lossless RGB images.
Transform domain techniques embed secret information in a
transform space of the signal means the process of embedding
data in the frequency domain of a signal to make them more
robust to attack such as adding noise, compression, cropping,
some image processing etc[10]. Many transform domain
variations exist. One method is to use the Discrete Cosine
Transformation (DCT) as a vehicle to embed information in
image. Another method would be the use of wavelet
transforms. Advantages of transform domain include higher
level of robustness against simple statistical analysis.
JSteg sequentially replaces the LSB of the non-zero quantized
DCT coefficients with secret message bits whereas in JPHide
the quantized DCT coefficients are not selected sequentially
but selected randomly by a pseudo-random number generator
[10]. F5 comes after a series of F3 and F4 [11]. F5
steganographic algorithm was introduced by Westfield and it
embeds message bits into randomly chosen DCT coefficients.
The F5 algorithm employs matrix embedding that minimizes
the necessary number of changes to hide a message of certain
length.
The discrete wavelet transform (DWT) method is favored over
the DCT method in steganography because DWT provides
better image resolution at various levels. DWT converts spatial
domain information to the frequency domain information and it
clearly partitions the high-frequency and low-frequency
information on a pixel by pixel basis [12]. Wavelets are
mathematical functions that divide data into frequency
components, which makes them ideal for image compression.
Haar wavelet transform is the widely used wavelet transform
due to the simplicity in implementation [13].
IV. METHODOLOGY
A. Algorithm for Embedding Process
1. Get cover image and secret image.
2. Calculate daubechies wavelet transform of the cover
image.
3. Encrypt the pixel value of secret image by XORing
pixel value with secret key.
4. Embed the encrypted pixel values of secret image
into the wavelet coefficients. This is done in
following method.
i. Select any two bands (i.e. HL/HH or LH/HH
or HL/LH).
ii. Get the secret information (image).
iii. Get lower four bits of secret information and
embed into the lower four bits of wavelet
coefficients of one of the band selected.
978-1-4673-6744-8/15/$31.00@2015 IEEE
iv. Get higher four bits of secret information
and embed into the lower four bits of
wavelet coefficients of other band selected.
5. Calculate inverse daubechies wavelet transform.
6. Stego image is generated.
B. Algorithm for Extraction Process
1. Get stego image.
2. Calculate daubechies wavelet transform of the stego
image.
3. Obtain the coefficients of the selected band where the
pixel value of secret image is hidden.
4. Extract the pixel values of secret image from wavelet
coefficients. The extraction is done in following
method.
i. Get coefficients from one of the selected
bands of wavelet.
ii. Obtain lower four bits of the coefficients.
iii. Get coefficients from other selected bands of
wavelet.
iv. Obtain lower four bits of the coefficients.
v. Combine this four bits with the four bits
obtained in (ii) by making this four bits as
higher bits and other four bits as lower bits.
vi. Decrypt the obtained value by XORing it
with secret key.
5. Secret image is generated.
V. EXPERIMENTS AND EVALUATIONS
A. Simulation Parameters
x Coding Platform: Matlab
x Cover and Secret Image: 24-bit color
x Cover Image size: 512 X 512
x Secret Image size: 128 X 128
B. Comparison of different bands of Daubechies wavelet
transform for embedding secret image
5 different secret images and 17 different cover images are
taken and secret image is embedded in different combination
of bands (horizontal/diagonal, vertical/diagonal,
vertical/horizontal) and performance is evaluated in terms of
mean square error and peak signal to noise ratio.
In table I, out of 17 tests, 12 give less mean square error and
hence more peak signal to noise ratio in horizontal/diagonal
band and 5 give less mean square error in vertical/diagonal
band.
In table II, out of 17 tests, 12 give less mean square error and
hence more peak signal to noise ratio in horizontal/diagonal
band, 4 give less mean square error in vertical/diagonal band
and 1 gives less mean square error in vertical/horizontal band.
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)

Table I: Calculation of MSE and PSNR at different bands with secret7.jpg as Table III: Calculation of MSE and PSNR at different bands with secret9.jpg as
secret image secret image

secret7.jpg (Secret Image) Secret9.jpg (Secret Image)

Cover horizontal/ vertical/ vertical/ Cover horizontal/ vertical/ vertical/
Image diagonal diagonal horizontal Image diagonal diagonal horizontal
MSE PSNR MSE PSNR MSE PSNR MSE PSNR MSE PSNR MSE PSNR
im 398.19 22.13 229.71 24.51 628.333 20.14 im 493.41 21.19 187.31 25.40 583.28 20.47
image1 172.50 25.76 321.57 23.05 1238.50 17.20 image1 153.20 26.27 323.17 23.03 1111.2 17.67
image2 203.70 25.04 182.54 25.51 696.405 19.70 image2 184.23 25.47 248.42 24.17 671.75 19.85
image3 209.78 24.91 237.52 24.37 505.039 21.09 image3 199.31 25.13 213.70 24.83 431.59 21.78
image4 233.73 24.44 325.88 23.00 1520.00 16.30 image4 342.02 22.79 399.94 22.11 1184.6 17.39
image5 131.90 26.92 284.68 23.58 973.316 18.24 image5 228.62 24.54 272.46 23.77 849.67 18.83
image6 228.98 24.53 244.36 24.25 365.103 22.50 image6 135.87 26.79 236.82 24.38 292.00 23.47
image8 203.79 25.03 260.07 23.98 473.548 21.37 image8 200.58 25.10 226.01 24.58 415.02 21.95
image1 158.25 26.13 183.42 25.49 338.497 22.83 image10 178.64 25.61 200.56 25.10 298.36 23.38
0
image1 174.96 25.70 298.32 23.38 21.54 205.53 25.00 241.21 24.30 334.27 22.89
455.523 image11
1
image1 170.90 25.80 297.51 23.39 21.79 166.28 25.92 179.05 25.60 311.04 23.20
430.382 image12
2
image1 190.31 25.33 291.06 23.49 1230.00 17.24 244.00 24.25 235.63 24.40 977.80 18.22
image13
3
image1 941.98 18.39 238.51 24.35 20.23 237.87 24.36 216.45 24.77 604.80 20.31
615.608 image14
4
image1 294.72 23.43 259.25 23.99 18.63 229.06 24.53 239.02 24.34 794.90 19.12
890.154 image15
5
image1 240.89 24.31 251.75 24.12 1038.50 17.96 196.78 25.19 214.62 24.81 988.31 18.18
image16
6
image1 221.87 24.67 243.39 24.26 21.98 190.74 25.32 203.19 25.05 275.73 23.72
411.758 image17
7
image1 178.68 25.61 164.12 25.97 23.66 180.91 25.55 157.77 26.15 240.35 24.32
279.510 image19
9

Table II: Calculation of MSE and PSNR at different bands with secret2.jpg as Table IV: Calculation of MSE and PSNR at different bands with secret6.jpg
secret image as secret image

secret2.jpg (Secret Image) secret6.jpg (Secret Image)

Cover horizontal/ Vertical/ vertical/ Cover horizontal/ Vertical/ vertical/
Image diagonal diagonal horizontal Image diagonal diagonal horizontal
MSE PSNR MSE PSNR MSE PSNR MSE PSNR MSE PSNR MSE PSNR
im 950.2 14.14 394.038 22.17 245.639 24.22 im 342.73 22.78 204.07 25.03 157.765 26.15
image1 732.47 19.48 872.841 18.72 3807.90 12.32 image1 225.92 24.59 461.11 21.49 1457.30 16.49
image2 419.87 21.90 595.430 20.38 2139.70 14.82 image2 223.39 24.64 214.33 24.82 1046.70 17.93
image3 637.30 20.08 730.619 19.49 1638.60 15.98 image3 233.86 24.44 260.53 23.97 662.007 19.92
image4 708.25 19.62 3984.40 12.12 4486.10 11.61 image4 254.38 24.07 879.46 18.68 1267.10 17.10
image5 528.33 20.90 791.091 19.14 3155.20 13.14 image5 206.28 24.98 318.35 23.10 1475.90 16.44
image6 504.87 21.09 809.075 19.05 816.448 19.01 image6 261.31 23.95 288.02 23.53 289.803 23.51
image8 483.70 21.28 384.771 22.27 1232.70 17.22 image8 234.55 24.42 221.62 24.67 487.715 21.24
image10 581.21 20.48 581.682 20.48 854.132 18.81 image10 186.07 25.43 230.54 24.50 323.423 23.03
image11 573.87 20.54 790.859 19.15 1127.50 17.61 image11 174.42 25.71 342.57 22.78 464.533 21.46
image12 505.74 21.09 628.611 20.14 889.407 18.64 image12 231.38 24.48 202.08 25.07 385.474 22.27
image13 493.96 21.19 814.702 19.02 3818.50 12.31 image13 173.44 25.73 299.10 23.37 1575.80 16.15
image14 214.81 24.81 778.628 19.21 2156.90 14.79 image14 227.33 24.56 377.95 22.35 1007.50 18.09
image15 512.30 21.03 776.525 19.22 2132.50 14.84 image15 200.11 25.11 272.25 23.78 1066.90 17.85
image16 596.62 20.37 583.599 20.47 2508.00 14.13 image16 271.59 23.79 329.48 22.95 1308.30 16.96
image17 485.81 21.26 480.759 21.31 720.347 19.55 image17 203.55 25.04 294.26 23.44 329.664 22.95
image19 715.71 19.58 515.019 21.01 675.622 19.83 image19 294.17 23.44 194.99 25.23 296.315 23.41

978-1-4673-6744-8/15/$31.00@2015 IEEE
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)

Table V: Calculation of MSE and PSNR at different bands with secret4.jpg as x Case I:- Secret Image: secret7.jpg
secret image
In this case, 14 cover images out of 17 images have less mean
square error in Daubechies wavelet transform than in Haar
Secret4.jpg (Secret Image) wavelet transform.
Cover horizontal/ Vertical/ vertical/
Image diagonal diagonal horizontal
MSE PSNR MSE PSNR MSE PSNR
im 162.27 26.02 243.17 24.27 227.559 24.56
image1 338.68 22.83 867.44 18.74 1593.80 16.10
image2 218.33 24.74 271.86 23.78 1069.40 17.83
image3 304.62 23.29 373.66 22.40 825.990 18.96
image4 332.22 22.91 453.26 21.56 1263.80 17.11
image5 272.03 23.78 386.66 22.25 1474.00 16.44
image6 216.54 24.77 230.78 24.49 357.139 22.60
image8 220.32 24.70 226.89 24.57 770.203 19.26
image10 308.57 23.23 302.59 23.32 400.939 22.10
image11 238.77 24.35 343.85 22.76 504.226 21.10
image12 268.73 23.83 315.11 23.14 458.779 21.51
image13 265.43 23.89 412.87 21.97 1717.00 15.78
image14 156.52 26.18 358.21 22.58 1082.10 17.78
image15 333.94 22.89 366.54 22.48 1029.30 18.00
image16 324.90 23.01 341.85 22.79 1231.60 17.22
image17 218.68 24.73 226.78 24.57 421.911 21.87
image19 256.95 24.03 249.36 24.16 199.966 25.12

In table III, out of 17 tests, 13 give less mean square error and
hence more peak signal to noise ratio in horizontal/diagonal
band and 4 give less mean square error in vertical/diagonal
band. x Case II:- Secret Image: secret2.jpg
In this case, 12 cover images out of 17 images have less mean
In table IV, out of 17 tests, 12 give less mean square error and square error in Daubechies wavelet transform than in Haar
hence more peak signal to noise ratio in horizontal/diagonal wavelet transform.
band and 5 give less mean square error in vertical/diagonal
band.

In table V, out of 17 tests, 15 give less mean square error and

hence more peak signal to noise ratio in horizontal/diagonal
band and 2 give less mean square error in vertical/diagonal
band.

In above five tables, out of 85 tests, 64 tests give less mean

square error in horizontal/diagonal band, 20 tests give less
mean square error in vertical/diagonal band and 1 test gives
less mean square error in vertical/horizontal band.
Horizontal/diagonal band has more images with less mean
square error and vertical/diagonal band also contains some
number of images with less mean square error. Hence, it can
be concluded that diagonal band contains less information of
an original image.
C. Comparison Between Haar and Daubechies Wavelet
Transform
5 different secret images and 17 different cover images are
taken and performance of Haar and Daubechies wavelet
transform is compared on these images in terms of mean
square error.

978-1-4673-6744-8/15/$31.00@2015 IEEE
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)
x Case III:- Secret Image: secret9.jpg
In this case, 12 cover images out of 17 images have less mean
square error of Daubechies wavelet transform than in Haar
wavelet transform.
x Case IV:- Secret Image: secret6.jpg
In this case, 14 cover images out of 17 images have less mean
square error in Daubechies wavelet transform than in Haar
wavelet transform.
978-1-4673-6744-8/15/$31.00@2015 IEEE
x Case III:- Secret Image: secret4.jpg
In this case, 13 cover images out of 17 images have less mean
square error in Daubechies wavelet transform than in Haar
wavelet transform.
In above 5 cases, a total of 65 out of 85 tests give less mean
square error in Daubechies wavelet transform than Haar
wavelet transform. Hence experiments showed that
Daubechies wavelet transform gives better quality of secret
image extracted from stego image.
VI. CONCLUSION
In steganography, secret information is to be embedded in
such a way that it does not make significant change to cover
file. In this research, Daubechies discrete wavelet transform
technique is used to embed the secret information and
different tests have been performed. Different results show
that diagonal band of wavelet transform carries less
information of original image and hence coefficients of this
band can be used to embed the secret information without
much change to an original image. Instead of taking single
band, embedding secret information in combination of bands
such as vertical/diagonal or horizontal/diagonal band gives
better result in terms of MSE.
Also the performance of Daubechies DWT and Haar DWT is
evaluated in terms of PSNR and MSE. Different tests show
that Daubechies DWT gives less MSE and hence more PSNR
compared to Haar DWT.
 2015 9th International Conference on Software, Knowledge, Information Management and Applications (SKIMA)

REFERENCES [11] P. Bateman and H. G. Schaathun, “Image Steganography and

Steganalysis”, Master’s thesis in Security Technologies & Applications,
[1] B. G. Banik and S. K. Bandyopadhyay, “A DWT Method for Image University of Surrey, 2008
Steganography”, International Journal of Advanced Research in [12] T. Morkel and J.H.P. Eloff, “An Overview Of Image Steganography”,
Computer Science and Software Engineering, 2013 Information and Computer Security Architecture (ICSA) Research
[2] A. Joseph and T. Narasimmalou, “Optimized Discrete Wavelet Transform Group, 2012
based Steganography”, IEEE Interational Conference on Advanced [13] H. Rohil et al., “Optimized Image Steganography using Discrete Wavelet
Communication Control and Computing Technologies, 2012 Transform (DWT)”, International Journal of Recent Development in
[3] Chi-Kwong Chan, L.M. Cheng, “Hiding data in images by simple LSB Engineering and Technology, 2014
substitution”, Pattern Recognition Society, 2004
[4] C. Simpson, "The Rules of Unified English Braille", Round Table on
Information Access for People with Print Disabilities Inc, Australia,
2010
[5] J. M. Guo and T. N. Le, “Secret Communication Using JPEG Double
Compression”, Signal Processing Letters IEEE, 2010.
[6] S. M. Karim et al., “A New Approach for LSB Based Image
Steganography using Secret Key”, International Conference on
Computer and Information Technology, 2011
[7] S. Tiwari and R. P. Mahajan, “A Secure Image Based Steganographic
Model Using RSA Algorithm and LSB Insertion”, International Journal
of Electronics Communication and Computer Engineering, 2012
[8] M. A. Ahmad et al., “Achieving Security for Images by LSB and MD5”,
Journal of Advanced Computer Science and Technology Research, 2012
[9] W. Luo et al., “Security Analysis on Spatial 1 Steganography for JPEG
Decompressed Images”, Signal Processing Letters IEEE, 2011.
[10] C. P. Sumathi and T. Santanam, “A Study of Various Steganographic
Techniques Used for Information Hiding”, International Journal of
Computer Science & Engineering Survey, 2013

APPENDIX

Cover Image

Wavelet Transform of an image

Original Secret Image Extracted Secret Image

Stego Image
978-1-4673-6744-8/15/$31.00@2015 IEEE