ON THE IMAGE QUALITY AND ENCODING TIMES OF LSB, MSB AND COMBINED LSB-MSB

International Journal of Computer Science & Information Technology (IJCSIT) Vol 7, No 4, August 2015
DOI:10.5121/ijcsit.2015.7407 79
ON THE IMAGE QUALITY AND ENCODING TIMES
OF LSB, MSB AND COMBINED LSB-MSB
STEGANOGRAPHY ALGORITHMS USING DIGITAL
IMAGES
Solomon O.Akinola and Adebanke A.Olatidoye
Department of Computer Science, University of Ibadan, Nigeria
ABSTRACT
The Least Significant Bit (LSB) algorithm and the Most Significant Bit (MSB) algorithm are steganography
algorithms with each one having its demerits. This work therefore proposed a Hybrid approach and
compared its efficiency with LSB and MSB algorithms. The Least Significant Bit (LSB) and Most
Significant Bit (MSB) techniques were combined in the proposed algorithm. Two bits (the least significant
bit and the most significant bit) of the cover images were replaced with a secret message. Comparisons
were made based on Mean-Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR) and the encoding
time between the proposed algorithm, LSB and MSB after embedding in digital images. The combined
technique produced a stego-image with minimal distortion in image quality than MSB technique
independent of the nature of data that was hidden. However, LSB algorithm produced the best stego-image
quality. Large cover images however made the combined algorithm’s quality better improved. The
combined algorithm had lesser time of image and text encoding. Therefore, a trade-off exists between the
encoding time and the quality of stego-image as demonstrated in this work.
KEYWORDS
Steganography, Cover image, Most Significant Bit (LSB), Least Significant Bit (LSB).
1. INTRODUCTION
Steganography is the art and science of hiding sensitive information in ways that prevent
detection. The purpose of steganography is to convey a message in such a way that nobody apart
from the sender and intended recipient suspects the existence of the message. These messages are
transferred through cover carriers such as text, audio, images and protocols [1, 2]. The secret
message could be a plaintext, cipher text or images. The embedding of the message into a cover
object results in the production of a stego-image. Images are mostly used as cover objects in
steganography.
Different image Steganography technique exists which are classified into spatial domain and
transform domain steganography. In spatial domain scheme, the secret information is directly
embedded. Its high capability of hiding and easy retrieval makes it to be used frequently. An
example is the least significant bit algorithm which is key to the embedding algorithm proposed
in this paper.
Transform domain scheme is used for hiding a large amount of data. It hides information in
frequency domain by altering magnitude of all transforms of cover image. Discrete
Cosineransform (DCT), Discrete Fourier Transform, and Wavelet Transform are the main types

80
of transforms used in steganography. These transforms all have coefficients associated with
them. The secret data is hidden within these coefficients which also defines how the image or file
should be transformed [3]. Examples include JPEG Steganography and Spread Spectrum.
The performance of a steganography technique can be measured using several parameters, among
which are imperceptibility, robustness and capacity. Imperceptibility is defined as the ability to
avoid detection, i.e. the inability to determine the existence of a hidden message. This makes it
an important requirement in steganography. Robustness refers to how well a steganography
technique can resist the extraction of hidden data. It measures the ability of the steganography
technique to survive the attempts of removing the hidden information. Such attempts include,
image manipulation (like cropping or rotating), data compression and image filtering [4]. Payload
Capacity represents the maximum amount of information that can be safely embedded and
retrieved in a work without being statistically detectable. When compared with watermarking
that requires embedding only a small amount of copyright information, Steganography requires
sufficient embedding capacity [5].
The Least Significant Bit (LSB) is one of most common embedding techniques. The least
significant bit is the least value in a binary number. In LSB algorithm, data is hidden in the least
significant bits of the cover image wfqhhich is not noticeable when viewed with the human eye
[4]. The most significant bit (also called the high-order bit) is the bit position in a binary number
having the greatest value.
The aim of this work is to compare the image quality and the encoding times of LSB, MSB and
the proposed Combined-LSB-MSB Steganography algorithms using digital images. The
objectives are to:
• Combine the LSB and MSB techniques into a Hybrid algorithm that embeds secret
message bits into the least significant bit and most significant bit of the cover image.
• Compare the LSB, MSB and the proposed algorithms (named Combined-LSB-MSB and
hence called Hybrid) in terms of encoding time, MSE (Mean Squared Error) and PSNR
(Peak Signal to Noise Ratio).
• Test the algorithms using different image formats (JPG and PNG) and the quality of image
with increase in file size.
The rest of the paper is organized as follows. Section II reviews existing image steganography
methods and section III presents the proposed image embedding method. The experimental
results & discussion are shown in section IV and conclusions are drawn in section V.
2. RELATED WORK
There has been several researches in hiding data inside an image using steganography technique.
In Warkentin et al. [6] proposed algorithm, the idea was to hide data inside the audiovisual files.
El-Emam’s [7] proposed steganography algorithm is based on hiding a large amount of data file
inside a coloured bitmap image. In his work, he filtered and segmented the image by using bits
replacement on the appropriate pixels. A concept defined by main cases with their sub cases for
each byte in one pixel was used to select these pixels randomly rather than sequentially. This
concept was both visual and statistical. The result of this concept was that 16 main cases with
their sub cases covered all aspects of the input data into color bitmap image. Three layers
provided high security which made it difficult to break through the encryption of the input data
and also undectable when steganalysis is applied. it was concluded that a large amount of data
that occupies 75% of the image size can be embedded efficiently and the output will be of high
quality.

81
Chen et al. [8] modified a method proposed by Chang et al. [9] using the side match method. In
this method, data was hidden in the edge portions of the image. The image quality was improved
while maintaining the same embedding capacity because the human eyes could rarely see
differences in the edge portion. The embedding capacity can also be adjusted based on the
demands of individual users. In addition to the improvement on image quality, the proposed
approach provided respectable security as well. Wu and Tsai [10] proposed an algorithm using
pixel-value differencing which partioned the original image into non-overlapping blocks of two
consecutive pixels. A different value was calculated from the values of the two pixels in each
block. All possible different values were classified into a number of ranges. The human vision
sensitivity to gray value variations from smoothness to contrast was used in selecting the range
intervals. A new value which replaced the different value was used to embed the value of a sub-
stream of the secret message. The width of the range that the different value belongs to
determines the number of bits that can be embedded in a pixel pair. However, in this method the
modification is never out of the range interval. The result produced by this method is more
imperceptible than those yielded by simple least significant bit replacement method. The secret
message that was embedded can be extracted from the resulting stego-image without making
reference to the method of the original cover image. The security of the method was shown using
dual statistics attack.
Scott [11] work on steganographic techniques using digital images used several iterations of
replacement strategies during the construction of the application. The aim was to implement a
replacement and extraction steganography scheme using cover images. To extract the embedded
textual information from the image, the image created by the application must be processed. This
processing outputs the original message and some extra erroneous information. Comparism
between LSB replacement scheme with MSB replacement scheme asserted that MSB produced
noticeable differences to the cover during the most significant bit replacement. Rohit and Tarun
[12] compared LSB and MSB based steganography in gray-scale images. It was concluded that
the resulting stego-image using LSB shows no distortion when compared with the original image.
The performance of LSB was better than that of MSB. Kanzariya and Nimavat [4] compared
various image steganography techniques. The objectives were to identify the requirements of a
good steganography algorithm and to determine steganography techniques that are suitable for
different applications. In this work, some criteria for imperceptibility of an algorithm were
proposed.
3. THE PROPOSED METHOD
The proposed hybrid algorithm combines the LSB and MSB Steganography techniques. Two bits
(the least significant and the most significant bits) of the cover images were replaced with a
secret message. Figure 1 shows the framework of the proposed algorithm.

82
Figure 1: Framework for the Proposed Hybrid-LSB-MSB Algorithm
From figure 1, once the cover image and the secret message (image or text) has been selected, the
embedding stage of the combined algorithm takes two bits of the secret message and embeds the
first message bit in the least significant bit of the cover image byte and the second message bit in
the most significant bit of the cover image byte. The output of this process is a stego-image. The
retrieving stage is just the inverse of the embedding stage.
A. Embedding Algorithm
Begin
Load the cover image
Convert image to byte array
Convert message data to byte array
If message cannot be contained in cover image
Exit with error message
Else
For each bit in the message byte
Begin
If LSB
Hide message bit in the lsb of the corresponding cover image byte
If MSB
Hide message bit in the msb of the corresponding cover image byte
If HYBRID
Get two message bits

83
Hide the first message bit in the lsb of the corresponding cover image byte
Hide the second message bit in the msb of the corresponding cover image byte
End
End
B. Decoding Algorithm
Begin
Load stego image
Convert stego image into byte array
If decoding type is LSB
Begin
For the first 32 byte
Copy the lsb into an array of length 32
Convert the array into integer value
Create an array of length of the integer value
Starting from length 32+1 of the stego-
image array
Begin
Copy the lsb of the equivalent stego array into an array of length 8
Convert the array into a byte value and save in the corresponding index of the
created array
Convert the array value into string or image
End
End
The same approach goes for MSB and HYBRID
End
C. File Format
Any image file format can be used as both the cover image and the secret image. However, the
image was first converted into PNG format before anything can be done on it. After the whole
process, the image was converted back to its original format. PNG format is preferred because it
is supported by the Java image IO library; it applies lossless file compression method and allows
for easy interchange and viewing of image data stored on local or remote computer systems [13].
Also, it seems to maintain a high degree of image quality after the message has been embedded
[11].
D. Comparison Procedures
To compare the image quality of the three algorithms i.e. the LSB, MSB and the proposed
Hybrid algorithm, three metrics were used, which are the Mean-Squared Error (MSE), the Peak
Signal-to-Noise Ratio (PSNR) and the encoding time.
• Mean-Squared Error (MSE)
The MSE represents the cumulative squared error between the cover image and the stego-image.
To calculate the mean-squared error (MSE) between two images I1 (M, N) and I2 (M, N) the
equation is as follows:

84
M and N are the number of rows and columns in the input images respectively [14].
• Peak Signal-to-Noise Ratio [PSNR]
The PSNR measures the statistical difference between the cover and stego image [15]. The mean-
squared error value is needed to compute the PSNR. The equation is as follows:
The value of R is 255.
However, the lower the MSE value and the higher the PSNR value then the better the quality of
the image.
4. RESULTS AND DISCUSSION
A simple system was developed to implement the LSB, MSB and the proposed combined
algorithms using JAVA programming language. There are two sides to the system, the encoding
interface and the decoding interface for hiding and retrieving purposes respectively.
We tested the system using two different image formats (roses.jpg, giraffe.png) as cover images.
A blue-footed booby bird (Figure 2) with dimension 160 x 120 pixels and file size of 4 kilo byte
was used as the message image for each cover image respectively. A 30.4 kilo byte document
was also used as message text.
Figure 2: Secret Image
In order to evaluate the performance of the proposed method, stego-images from the LSB, MSB
and the proposed Hybrid method were compared using MSE, PSNR and encoding time metrics
The methods were also tested with increased sizes of the images. Figures 3a, 3b, 4a and 4b show
the differences between the Least Significant Bit (LSB), Most Significant Bit (MSB) and the
combined algorithm after embedding messages in them.

85
Figure 3a (560 x 448 pixels rose.jpg hiding an image): (I) Original image (II) Stego-image using LSB
Figure 3a (560 x 448 pixels rose.jpg hiding an image): (III) Stego-image using MSB (IV) Stego-image
using Combined LSB-MSB
Using roses.jpg with dimension 560 x 448 pixels as cover image and the blue-footed booby bird
image as message, it can be seen from image II of figure 3a that there are no noticeable
differences between the original cover image and the resultant image after hiding in the Least
Significant Bit. Images 3a (III) and 3a (IV) show noticeable differences when compared to the
original cover image using Most Significant Bit and combined LSB-MSB algorithms
respectively. However, MSB (3a III) shows much difference.
Increasing the dimension of roses.jpg to 5040 x 4032 pixels, the payload capacity increases for
MSB and proposed algorithm (figure 3b (III & IV). Therefore, the larger the cover image the
more data that can be stored.

86
Figure 3b (5040 x 4032 pixels rose.jpg hiding an image): (I) Original image (II) Stego-image using LSB
Combined LSB-MSB
Figure 3b (5040 x 4032 pixels rose.jpg hiding an image): (III) Stego-image using MSB (IV) Stego-image
Figure 4a shows the output of the newly created stego-images after hiding text with a file size of
30.4kb (31,160 bytes) in an image in PNG format. The dimension of the cover image, giraffe.png
is 750 x 1125 pixels. Image 4a (II) showed no noticeable difference when compared to the
original cover image after embedding text using LSB algorithm. The differences are noticeable at
the top sections of Figures 4a III and IV for MSB and the combined algorithms respectively.
Figure 4a (750 x 1125 pixels giraffe.png hiding text): (I) Original image (II) Stego-image using LSB

87
Figure 4a (750 x 1125 pixels giraffe.png hiding text): (III) Stego-image using MSB (IV) Stego-image using
Combined LSB-MSB
Increasing the dimension of the PNG file to 6750 x 10125 pixels produced a stego-image
indistinguishable from the original cover image when viewed with the human eyes for the three
algorithms (Figure 4b).
Figure 4b (6750 x 10125 pixels giraffe.png hiding text): (I) Original image (II) Stego-image using LSB
Figure 4b (6750 x 10125 pixels giraffe.png hiding text): (III) Stego-image using MSB (IV) Stego-image

88
5. HELPFUL HINTS
Table 1 shows the MSE, PSNR and encoding times of the cover images for image and text
embedding. It can be seen that a lower MSE value and a higher PSNR value for LSB algorithm
as compared to the MSB and proposed Hybrid algorithms for both image and text were obtained.
This results into a better image quality since the lower the MSE value and the higher the PSNR
value, the better the quality of the image and hence imperceptibility is improved.
Although the embedding capacity of the proposed method (Hybrid) is low compared to LSB, the
proposed method gives better performance in all the parameters than MSB. The stego-image
generated after embedding the secret message in the cover image is almost identical to the
original image. However, when the sizes of the cover images were increased, the image quality
of the proposed algorithm increased, which means that the larger the cover image, the better the
hiding capacity. Also, the encoding times of the proposed Hybrid algorithm for various sizes of
the different images were lesser compared to other methods.
Table 1: Values of Encoding Time, MSEs and PSNRs of stego-images in which image and text is embedded
respectively.
Figures 5a and 5b shows the bar chart of results obtained with rose.jpg of sizes 560 x 448 and
5040 x 4032 pixels after embedding an image using LSB, MSB and Hybrid algorithms
respectively. The LSB algorithm had the lowest MSE and highest PSNR values while the
proposed combined algorithm had the lowest encoding time.

89
Figure 5a: The Encoding Time (ms), MSE (dB) and PSNR (dB) of 560 x 448 pixels rose.jpg
Figure 5b: The Encoding Time (ms), MSE (dB) and PSNR (dB) of 5040 x 4032 pixels rose.jpg
Figures 6a and 6b shows the bar chart of results obtained with giraffe.png of sizes 750 x 1125
and 6750 x 10125 pixels after embedding text using LSB, MSB and com algorithms respectively.
The LSB algorithm also had the lowest MSE value and the highest PSNR value while the
proposed Hybrid algorithm had the lowest encoding time value.
Figure 6a: The Encoding Time (ms), MSE (dB) and PSNR (dB) of 750 x 1125 pixels giraffe.png

90
Figure 6b: The Encoding Time (ms), MSE (dB) and PSNR (dB) of 6750 x 10125 pixels giraffe.png
Overall, LSB gives a best performance in terms of MSE and PSNR than the MSB and the
proposed Hybrid algorithm while the proposed Hybrid algorithm gives the best performance in
terms of the encoding time and better than MSB. The result obtained in this work confirms the
submission of Rohit and Tarun [12] that LSB steganography is much better than MSB
steganography for hiding messages. Scott [11] paper also compared LSB replacement scheme
with MSB replacement scheme and asserted that MSB produced noticeable differences to the
cover during the most significant bit replacement.
6. CONCLUSION
In this work, a Hybrid (LSB-MSB) algorithm was developed for embedding images and text into
images. The Hybrid algorithm suggests the embedding of secret message bits into the least
significant bit and the most significant bit of the cover image. The performance measure of
image quality due to embedding for LSB, MSB and the proposed Hybrid technique was
evaluated based on three error metrics; Mean-Squared Error (MSE), Peak Signal-to-Noise Ratio
(PSNR) and the time it takes each algorithm to embed a message in a cover image. The results
presented and analyzed show that the stego-images of LSB have the highest PSNR and the lowest
MSE values, making it very efficient to hide the data inside the image. However, based on the
encoding time, the combined LSB-MSB algorithm takes lesser time in embedding than LSB and
MSB.
Conclusively, LSB algorithm gives a better performance than the combined LSB-MSB algorithm
but a larger file size of the cover image makes the combined algorithm produces images with
good quality. Nevertheless, the proposed Hybrid algorithm produced better image quality than
MSB algorithm.
In the future work, the security of using the hybrid algorithm could be improved by working on
the compression ratio for stronger embedding procedures and also the use of the proposed
combined LSB-MSB algorithm on large sized gray-scale images.

91
REFERENCES
[1] Currie, D.L. & Irvine, C.E. (1996). Surmounting the Effects of Lossy Compression on Steganography,
19th National Information Systems Security Conference, Oct 1996, Baltimore, Md., USA, pp. 194 -
201
[2] Jonathan Blackledge and Oleksandr Iakovenko (2014). Resilient Digital Image Watermarking for
Document Authentication, IAENG Internation Journal of Computer Science, 41 (1).
[3] Jessica Codr (2009). Unseen: An Overview of Steganography and Presentation of Associated Java
Application C-Hide, Washington University in St.Louis, Department of Computer Science &
Engineering, Bryan Hall, Campus Box 1045, One Brookings Drive, St. Louis, Missouri 63130,
http://www.cse.wustl.edu/~jain/cse571-09/ftp/stegano/index.html
[4] Kanzariya Nitin K. and Nimavat Ashish V. (2013). Comparison of Various Images Steganography
Techniques, International Journal of Computer Science and Management Research, Vol. 2, Issue 1
[5] Nagham Hamid, Abid Yahya, Badlishah Ahmad R. and Osamah Al-Qershi M. (2012). Image
Steganography Techniques: An Overview, International Journal of Computer Science and Security
(IJCSS), Vol. 6, Issue 3
[6] Warkentin M., M.B. Schmidt and Bekkering E. (2008). Steganography and Steganalysis, Intellectual
Property Protection for Multimedia Information Technology, Chapter XIX, pp. 374 - 380
[7] El-Emam N. N (2007). Hiding a Large Amount of Data with High Security Using Steganography
Algorithm, Journal of Computer Science, Vol. 3, pp. 223 - 232
[8] Chen P.Y., Wu W.E. (2009). A Modified Side Match Scheme for Image Steganography, International
Journal of Applied Science & Engineering, Vol. 7 pp. 53 – 60
[9] Chang C.C. and Tseng H.W. (2004). A Steganographic Method for Digital Images using Side Match
Pattern Recognition Letters, 1431-1437
[10] Wu P.C, Tsai W.H. (2003). A Steganographic Method for Images by Pixel-value Differencing, Pattern
Recognition Letters, pp. 1613 - 1626
[11] Scott Bishop (2004). Steganographic Techniques using Digital Images, California State University,
Hayward
[12] Rohit Garg and Tarun Gulati (2012). Comparison of Lsb & Msb Based Steganography in Gray-Scale
Images, International Journal of Engineering Research & Technology (IJERT), Vol. 1, Issue 8
[13] Lokeswara Reddy V., Dr. Subramanyam A., Dr. Chenna Reddy P. (2011). Implementation of LSB
Steganography and its Evaluation for Various File Formats, International Journal Advanced
Networking and Applications, Vol. 02, Issue 05, pp. 868 – 872
[14] Stuti Goel, Arun Rana & Manpreet Kaur (2013). A Review of Comparison Techniques of Image
Steganography, Global Journal of Computer Science and Technology Graphics & Vision, Vol.13, Issue
4, Version 1.0
[15] Kumar Shiva K.B., Raja K.B., Chhotaray R.K and Sabyasachi Pattnaik (2011). Performance
Comparison of Robust Steganography Based on Multiple Transformation Techniques, International
Journal Computer Technology Application, Vol. 2 (4), pp. 1035 – 1047.
Authors
Akinola Olalekan is a Senior lecturer of Computer Science at the University of Ibadan,
Nigeria. He had his PhD Degree in Software Engineering from the same University in
Nigeria. His research focus is on software quality assurance techniques.
Adebanke Olatidoye had her Masters Degree in Computer Science from University of
Ibadan, Nigeria. She also had her first Degree in Computer Science from Ladoke Akintola
University of Technology, Ogbomoso, Nigeria.

ON THE IMAGE QUALITY AND ENCODING TIMES OF LSB, MSB AND COMBINED LSB-MSB

More Related Content

What's hot (18)

Viewers also liked (17)

Similar to ON THE IMAGE QUALITY AND ENCODING TIMES OF LSB, MSB AND COMBINED LSB-MSB (20)

Recently uploaded (20)

ON THE IMAGE QUALITY AND ENCODING TIMES OF LSB, MSB AND COMBINED LSB-MSB