An enhanced difference pair mapping steganography method to improve embedding capacity

IJRET: International Journal of Research in Engineering and Technology eISSN: 2319-1163 | pISSN: 2321-7308
__________________________________________________________________________________________
Volume: 03 Special Issue: 07 | May-2014, Available @ http://www.ijret.org 196
AN ENHANCED DIFFERENCE PAIR MAPPING STEGANOGRAPHY
METHOD TO IMPROVE EMBEDDING CAPACITY
V.Poongavanam1
, L. Mary Shamala2
1
Master of Engineering, Computer Science and Engineering, IFET College of Engineering, Tamilnadu, India
2
Associate Professor, Computer Science and Engineering, IFET College of Engineering, Tamilnadu, India
Abstract
Reversible Data Hiding (RDH) aims to embed secret message into a cover image by slightly modifying its pixel values, and unlike
data hiding, the embedded message as well as the cover image should be completely recovered from the marked content. Many
methods have been implemented in RDH to maintain secure confidential data, such as Difference Expansion (DE), Histogram
Modification and Difference Pair Method (DPM). But its embedding capacity is quite low and also its PSNR value is low, then it takes
more time to complete the process. In order to overcome these issues, Modified Difference Pair Mapping (MDPM) technique of RDH
is proposed for embedding extra information in an image by making small modifications to its pixels. MDPM is the process of
dividing an image into blocks. Then divide the blocks into embeddable area and non-embeddable area. Embeddable area is used to
embed a secret message, and non-embeddable area is used to record the side information. In embeddable area, Centre pixel of 3x3
blocks is chooses as reference pixel, remaining 8 pixels is referred to as non-reference pixels. Then calculate difference value in non-
reference pixels. Finally difference histogram has been generated. In that difference histogram, peak point and zero point is obtained.
From this peak point and zero point of difference histogram, secret message is embedding. Location map and side information is used
in non-embeddable area to extract a secret message and original image. Theoretical analysis indicates that the proposed technique
can provide good trade-off between embedding capacity and stego-image quality.
Keywords: Difference Value calculation, Local complexity, Reversible Data Hiding, Reference and Non-reference
pixels, Histogram generation, Histogram Pair, Smooth region and Complex region.
----------------------------------------------------------------------***----------------------------------------------------------------------
1. INTRODUCTION
The hiding of data is frequently called steganography.
Steganography is a technology that hides a message within an
object. In that steganography, Reversible Data Hiding (RDH)
is the one of the scheme to hide a secret data. It is also same as
like data hiding, but unlike data hiding, RDH can recover the
embedded message as well as original image without any
distortion. Nowadays, RDH can be preferably used to hiding a
secret message. Generally, the performance of a RDH scheme
is evaluated by the capacity-distortion behavior. For achieving
a high embedding capacity (EC) and to obtain a good marked
image quality we have to reduce the embedding distortion as
much as possible. RDH Scheme has been used to hide secret
data into an image. Many approaches have been implemented
in RDH to provide secure data such as difference expansion,
histogram modification and Difference pair. The first
histogram-based RDH method is the Histogram Shifting [1].
This method uses peak and minimum points of the pixel-
intensity-histogram to embed data. It changes each pixel value
at most by 1, and thus a good marked image quality can be
obtained. However, its EC is quite low and this method does
not work well if the cover image has a flat histogram. To
facilitate it, Lee et al. [2] proposed to utilize the difference-
histogram instead. This method exploits the correlation among
neighboring pixels and can embed larger payload with reduced
distortion compared with Ni et al.’s. Lee et al.’s method can
be in fact implemented, in an equivalent way, by modifying
the two-dimensional pixel-intensity-histogram according to a
pixel-pair-mapping (PPM) which is an injective mapping
defined on pixel-pair.
In Difference Pair Mapping method, LSB algorithm is used
for embedding the data. But these methods are consuming
more time to complete the process, also PSNR value is low.
This paper presents a new method Modified Difference pair
mapping (MDPM) based on RDH scheme to achieve a high
embedding capacity and better PSNR value.
We remark that, RDH methods generally contain two basic
steps:
• Histogram generation: First, each local image region
consisting of several pixels i.e. pixel-pair is projected to a one-
dimensional space to get a difference value of pixel-pair. One-
dimensional histogram (e.g., difference- histogram) is
generated by counting the number of difference pair value.
• Histogram modification: Finally, embed data into the cover
image by modifying the histogram. In most cases, the
histogram bins with high frequencies are expanded to carry
data while some others are shifted to ensure the reversibility.

__________________________________________________________________________________________
By using the two-dimensional difference-histogram and this
specific DPM, compared with the conventional histogram
based methods, more pixels are used for carrying data while
the number of shifted pixels is reduced as well, and thus an
improved embedding performance is achieved.
The rest of the paper is organized as follows. The related
works are briefly introduced in Section II. Section III presents
the proposed RDH scheme in details. The comparisons with
the prior works are shown in Section IV. Section V concludes
this paper.
2. RELATED WORKS
Difference Expansion is the process of dividing the image into
pair of pixels and a secret message was embedded into the
difference between the pixels. DE first selects the adjacent
pixels, and then expands the difference values of two pixels to
embed the secret message. The secret message was embedded
using a location map which contains the location of data
carrying pixels. Using this difference expansion method can
achieve sufficient embedding capacity. But Difference
Expansion method suffers from low compressibility of
location image and lack of embedding capacity control.
Location map is utilized to record the pixels that may cause
the overflow problem. This method provides less embedding
capacity and allows limited bits of data to embed.
To overcome this issue, RDH introduced Histogram
modification method to embed a data into an image based on
pixel pair method. Histogram Modification is a process of
scanning the image and selecting the pixel pair, and then
calculating a difference value to form histogram. Then find a
peak point in that difference value to embed a data. This
method is used to provide lossless large volume data hiding
method. Embedding and extraction process is based on the
difference and pixel value.
Difference value for pixel-pair:
xi ,if i=0 (2.1)
 xixi 1 | , otherwise 0
Then have to select peak point in those pixel pair value. In that
peak point, if di>P, shift each pixel by 1 unit. Then,
Yi=
𝑥𝑖, 𝑖𝑓 𝑖 = 0
𝑥𝑖+1, 𝑖𝑓 𝑑𝑖 > 𝑃, 𝑥𝑖 > 𝑥𝑖 − 1
𝑥𝑖−1, 𝑖𝑓 𝑑𝑖 > 𝑃, 𝑥𝑖 < 𝑥𝑖 − 1
(2.2)
Yi= embedded value of pixel i.
di=P modify the pixel xi, according to the message bit b.



bix if xi>xi-1
 bix  if xi<xi-1 (2.3)
Based on these difference value and peak point value
embedding has to be done here. If difference value and peak
point value should be equal means, it should modify the pixel
value by message bit b.
2.1 Embedding Process
Step 1: First, it scans the image.
Step 2: Calculate the pixel difference (di), between adjacent
pixels (Xi, Xi-1) and generate the histogram.
Step 3: Determine the peak point (P), from pixel differences.
Step 4: If the di> P means, it shifts each pixel by 1 unit.
Step 5: If di=P means, it modify the pixel Xi according to
message bit b, by embedding this message bit into pixel value.
2.2 Extraction Process
Step 1: Extraction is the reverse process of this embedding
process, then it restores the pixel value to extract a secret
message from the cover image without any distortion.
Step 2: Scan the image in the same order as in the embedding
phase.
Step 3: Using the difference pixel (di) values and peak point
(P) values it restores the pixel in the stego image to extract a
secret data.
Step 4: Extract a secret data from cover image without any
distortion using pixel restoration.
This method provides better embedding capacity, But this
method will not work, if the difference value produces flat
histogram.
To overcome this issue, Difference Pair Mapping has been
proposed in RDH method. DPM is the process of dividing an
image into a pixel, and then it selects an adjacent pixel to form
a pixel-pair. Using this pixel pair by considering a pixel-pair
and its context, a local image region is projected to a two-
dimensional space to obtain a sequence consisting of
difference-pairs. Then, a two-dimensional difference-
histogram is generated by counting the difference-pairs.
Finally, reversible data embedding is implemented according
to a specifically designed difference-pair-mapping (DPM).
DPM is a natural extension of expansion embedding and
shifting techniques used in current histogram-based methods.
By using the two-dimensional difference-histogram and this
specific DPM, compared with the conventional histogram
based methods, more pixels are used for carrying data while
the number of shifted pixels is reduced as well, and thus an
improved embedding performance is achieved.

__________________________________________________________________________________________
For a pixel-pair(x, y), to compute difference values 𝑑1 =
𝑥 − 𝑦 and 𝑑2 = 𝑦 − 𝑧 to form a two-dimensional
difference-histogram of, where z is a prediction of y.
In this situation, since pixel-pair (x, y) has four modification
directions, the difference-pair (d1, d2) also has four
modification directions.
For example, by modifying y to y+1, the modification
direction to (x, y) is “up” and the corresponding modification
direction to (d1, d2) is “upper-left”, since d1 changes to d1 -1
and d2 changes to d2+1. Based on these four modification
directions, DPM have been implemented. But in this DPM
method based RDH, only one pixel of a pixel-pair should be
modified by one bit. So this method provides low embedding
capacity.
Our method is motivated by the aforementioned works and
will be described in details in next section.
3. PROPOSED WORK
RDH scheme introduces DPM method to embed a data into an
image. This Section is organized as follows. First in Section
3.1 Modified DPM-based scheme. In this way, by extending
difference- histogram (or, in an equivalent way, two-
dimensional pixel-intensity-histogram) to two-dimensional
difference-histogram, the proposed scheme extends current
histogram-based RDH methods. The detailed embedding and
extraction procedures of this scheme are summarized in
Section3.2 and section 3.3. Finally, Experimental results are
explained in Section 3.4.
3.1 Modified DPM based Scheme
MDPM is the process of dividing an image into embeddable
area and non-embeddable area. Fig.1 explains the overall
process of MDPM. The embeddable area is used for
embedding secret data and the non-embeddable area is used to
record the information of location map and side information.
To embed a secret data, first divide an embeddable area into
3x3 blocks. For each block, the center pixel is selected as the
reference pixel, and the other eight pixels are referred to as
non-reference pixels.
Fig.1. Architecture diagram
Then calculate a difference value for those non-reference
pixels. Using this difference value of non-reference pixels
histogram is generated. From that histogram, pairs of zero
point and peak point is obtained. Using this pair, secret data is
embedded into difference value. Finally, reversible data
embedding is implemented according to a specifically
designed MDPM. Then to extract a secret message from
original image location map has been used.
3.1.1 Local Complexity and Selection of Image
Region
The Cover image I is divided into block with the size of 3x3
pixels. Then select the embeddable area and non-embeddable
area using the measurement of the local complexity.
Fig.2. Image block with reference pixel Ci , satellite reference
pixel and non-reference pixels.
Centre pixel Ci and Cu
, Cd
, Cr
, Cl
and 1, 2, 3, 4, 5, 6, 7, 8
mentioned in Fig.2 reffered as reference pixel, satellite
reference pixels and non-reference pixels respectively. Local
Complexity is measured by following equation (3.1)

__________________________________________________________________________________________
iLC =Range (Ci
l
, Ci
r
, Ci
u
, Ci
d
) (3.1)
Where the Range (Ci
l
, Ci
r
, Ci
u
, Ci
d
) function is used to measure
the maximum and minimum values of four satellite reference
pixels Ci
l
, Ci
r
, Ci
u
, Ci
d
. If the complexity value is less than the
threshold value T, the current processing block is located in
smooth region, otherwise the block is located in complex
region. Smooth region is used here to embed secret data.
Complex region is used here to carrying location map and side
information.
3.1.2 Calculating Difference Value and Histogram
Generation
Smooth block is used to embedding a secret data, compared to
complexity region smooth block provide more embedding
capacity. Fig 3.a and b shows the process of pixel block and
difference block. Then smooth block non reference pixel is
used to embed a data. First have to select reference and non-
reference pixels in smooth block. Then have to calculate
difference value for those non-reference pixels.
Fig.3.a. Pixel blocks Fig.3.b. Difference block
The difference histogram of the non-reference pixels is
constructed. Then, two pairs (PP1, ZP1) and (PP2, ZP2) of the
peak and zero points of the histogram are obtained. Based on
these pairs, embedding should be processed in difference
block.
3.2 Data Embedding Phase
During the embedding process, the histogram shifting
technique is used in the proposed scheme. Fig.4 shows the
embedding process. Based on the difference value of non-
reference pixels histogram have been generated. From that
histogram pair (peak point and zero point) value embedding
should be processed here.
Fig.4 Embedding Process
The embedding process should be explained in the following
steps:
Step 1: The cover image I is partitioned into two areas, i.e.,
embeddable area and non-embeddable area.
Step 2: The embeddable area of the cover image is divided
into blocks with sizes of 3x3 pixels.
Step 3: For each block Bi€ S-block, compute the difference
value di of the non-reference pixels in Bi with equation-1
d1= P1- P2
d2=P2 -Ci
d3= P3-P4
d4 = P4-Ci (1)
d5 = P5-P6
d6 = P6-Ci
d7 = P7-P8
d8 = P8-Ci
Step 4: The difference histogram of the non-reference pixels is
constructed. Then two pairs, i.e., (PP1, ZP1) and (PP2, ZP2) of
the peak and zero points of the histogram are obtained.

__________________________________________________________________________________________
Step5: Scan the difference values di sequentially. If di=PP1 or
di=PP2, a secret bit can be embedded. Otherwise no difference
secret bit can be embedded.
𝑑𝑖′
=
𝑑𝑖 𝑖𝑓 𝑆 = 0
𝑑𝑖−1 𝑖𝑓 𝑠 = 1 𝑎𝑛𝑑 𝑑𝑖 = 𝑃𝑃1
𝑑𝑖+1 𝑖𝑓 𝑆 = 1 𝑎𝑛𝑑𝑑𝑖 = 𝑃𝑃2
(2)
Step6: Repeat step 5 until the embedded data S are embedded
completely.
These are steps to complete the embedding process, after
finishing these steps stego image is obtain. Then extraction
process will be continued to recover the cover image as well
as secret message.
3.3 Data Extraction Phase
In the extraction phase, the extra information, including
location map and side information should be reconstructed.
Non-embeddable area is used here to record the information of
location map and side information. To reconstruct he cover
image, the LSBs of the pixels in the non-embeddable area are
recovered.
The extraction phase should be explained in the following
steps:
Step 1: The stego image I′′ is divided into two areas, i.e., the
embeddable area and the non-embeddable area, in the same
manner used in the embedding phase.
Step 2: Divide the embeddable area of the stego image into
blocks with sizes of 3x3 pixels.
Step3: Compute the difference values of the non-reference
pixels in B′i using equation 1.
Step4: Scan the difference values d′i sequentially by using the
same order that was used in the data embedding phase.
Step5: The original difference value di can be reconstructed
using equation 3
𝑑𝑖 =
𝑑′𝑖+1 𝑖𝑓 𝑍𝑃1 < 𝑑𝑖 < 𝑃𝑃1
𝑑′𝑖−1 𝑖𝑓 𝑃𝑃2 < 𝑑𝑖 < 𝑍𝑃2
𝑑′𝑖 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(3)
Step 6: Repeat step 5 until all embedded data S have been
extracted.
After finishing these steps stego image and secret message is
recovered. To reconstruct the cover image I, the LSBs of the
pixels in the non-embeddable area are recovered according to
the embedded data S.
3.4 Experimental Results
DPM based on LSB algorithm has been given below in table1,
based on these results proposed work have been improve to
overcome these issues in their work.
Based on these results proposed work have been implemented.
Fig.5 shows the example image to calculate the embedding
capacity.
Table.1. Embedding Capacity
S.no DE (Embedding
Capacity in %)
Histogram
modification
(Embedding
Capacity %)
DPM using
LSB
(Embedding
Capacity in
%)
1. 39.4444 40.123 41.1111
2. 30.1133 32.2233 33.27160
MDPM method is used to provide high embedding capacity
and high PSNR value.
4. CONCLUSIONS
Reversible Data Hiding is a process of embedding a secret
data into an image, which allow exact recovery of the original
image by extracting secret message from the marked image
without any distortion. Existing RDH methods like Difference
Expansion, Histogram Modification and Difference Pair
Mapping method provides less image quality. In order to solve
this issue, a better method is proposed in this project to
improve the image quality. Theoretical analysis reveals that
MDPM method can exploit the better image redundancy and
achieve sufficient embedding capacity.
ACKNOWLEDGMENTS
The authors wish to thank Principal Dr.G.Mahendiran, Dean
Academics Professor S.Matilda and Associate Professor
L.Mary Shamala for giving the valuable suggestion and
support.

__________________________________________________________________________________________
REFERENCES
[1] Weiming Zhang, Xinlu Gui , Xiao long Li, , and Bin
Yang , “Reversible data hiding by two dimensional
difference histogram”, IEEE Transactions On
Information Forensics And Security, Vol. 8, NO. 7,
JULY 2013.
[2] C. C. Chang ,W. L. Tai, C. M. Yeh, and, “Reversible
data hiding based on histogram modification of pixel
differences,” IEEE Trans. Circuits Syst. Video
Technol., vol. 19, no. 6, pp. 906–910, Jun. 2009.
[3] V. Fotopoulos ,E. Varsaki, , A. N. Skodras, “A
reversible data hiding technique” in Proc Technical
Report HOU-CS-TR-2006-08-GR. A. M.
[4] Alattar, “Reversible watermark using the difference
expansion of a generalized integer transform,” IEEE
Trans. Image Process., vol. 13, no. 8, pp. 1147–1156,
Aug. 2011.
[5] Y. Hu, H. K. Lee, and J. Li, “DE-based reversible data
hiding with improved overflow location map,” IEEE
Trans. Circuits Syst. Video Technol., vol. 19, no. 2, pp.
250–260, Feb. 2009.
[6] Nag ham Hamid, Abid Yahya, R. Badlishah Ahmad &
Osamah M. Al-Qershi, “Image Steganography
Techniques: An Overview”, International Journal of
Computer Science and Security (IJCSS), Volume (6) :
Issue (3) : 2012.
[7] Shashikala Channalli, Ajay Jadhav,”Steganography An
Art of Hiding Data” International Journal on
Computer Science and Engineering Vol.1(3), ISSN :
0975-3397 2009.
[8] Hsien-Wei Yang1,2, I-En Liao And Chaur-Chin Chen,
“Reversible Data Hiding Based on Median Difference
Histogram”, International journal of information
science and engineering 27, 577-593 (2011).
[9] Yun Q. Shi “Statistical Modeling in Reversible Data
Hiding”, International Journal of Computer Science
and Security (IJCSS),2011.
[10] Ching-Yu Yang & Wu-Chih Hu, “Reversible Data
Hiding in the Spatial and Frequency Domains”,
International Journal of Image Processing, Volume
(3):Issue (6), 2011.
[11] A. E.Mustafa A.M.F.ElGamal M.E.ElAlmi Ahmed.BD,
“A Proposed Algorithm for Steganography in Digital
Image Based on Least Significant Bit”, Research
Journal Specific Education, Issue No. 21, April. 2011.
[12] Chang-Chun Chan, Chin-Chan Chang,” Hiding Secret
Information Lossless Compression codes”,
International Journal of Innovative Computing, vol.7,
no.10, October 2011, ISSN 1349-4198.
[13] Yun Q. Shi1, Zhicheng Ni, Dekun Zou, Changyin
Liang and Guorong Xuan, “Lossless data hiding:
Fundamentals, algorithms and applications”, ISCAS,
vol.4, 2004.
[14] Hyoung Joong Kim, Vasiliy Sachnev, and Dong Hoi
Kim, “New reversible data hiding algorithm based on
difference expansion method ”, vol.7, no.5 2005.
[15] Vinit Agham and Tareek Pattewar1, “A survey on
Seperable Reversible Data hiding”, IMACST: Volume
4 Number 1 May 2013 1857-7202.
BIOGRAPHIES
Poongavanam.V(1991) received the degree
B.Tech,IT from Anna University, Chennai in
2012. She currently pursues Master of
Engineering, in Computer Science under the
same University. Her research interests
include Steganography, Image Processing and Data Mining.
Mary Shamala.L has pursued M.Tech degree
with specialization in Information security
from Pondicherry Engineering College,
Puducherry, India. She is currently an
Associate Professor in IFET College of
Engineering, Tamilnadu, India. Her research
interests include network and information security, privacy
enhancing technologies in social computing and lightweight
cryptography.

An enhanced difference pair mapping steganography method to improve embedding capacity

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to An enhanced difference pair mapping steganography method to improve embedding capacity (20)

More from eSAT Publishing House (20)

Recently uploaded (20)

An enhanced difference pair mapping steganography method to improve embedding capacity