An improved robust and secured image steganographic scheme

INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATION
International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN
0976 – 6464(Print), ISSN 0976 – 6472(Online) Volume 3, Issue 3, October- December (2012), © IAEME
ENGINEERING & TECHNOLOGY (IJECET)

ISSN 0976 – 6464(Print)
ISSN 0976 – 6472(Online)
Volume 3, Issue 3, October- December (2012), pp. 22-33 IJECET
© IAEME: www.iaeme.com/ijecet.asp
Journal Impact Factor (2012): 3.5930 (Calculated by GISI) ©IAEME
www.jifactor.com

AN IMPROVED ROBUST AND SECURED IMAGE
STEGANOGRAPHIC SCHEME

Nagham Hamid1, Abid Yahya2, and R. Badlishah Ahmad3, Osamah M. Al-Qershi4
1, 2, 3
(Communication and Computer Engineering School, University Malaysia Perlis,
Perlis, Malaysia)
4
(School of Electrical and Electronic Engineering, University of Science Malaysia, Penang,
Malaysia)

ABSTRACT

Due to the nature of the current digital world, many techniques have become essential for the
protection of secret data. The protection of such secret information has led to the
development of different kinds of techniques in different categories. Of all of these,
steganography has the advantage of concealing vital information in an imperceptible manner.
An improved steganographic system is presented in this paper, which successfully embeds
secret data within the frequency domain by modifying the Discrete Cosine Transformation
(DCT) coefficients. Based on selection criteria, certain blocks are selected for the
concealment of data. To ensure a full recovery for the hidden message, an embedding map is
proposed to indicate the selected embedding blocks. To secure the embedding map, Speed-
Up Robust Features (SURF) is used to dynamically define the locations in which the
embedding map is concealed. In addition, the embedding map is hidden in the frequency
domain as well by modifying the Discreet Wavelet Transformation (DWT) coefficients in a
content-based manner. The obtained results show the robustness of the proposed system
against Additive White Gaussian Noise (AWGN) and JPEG compression attacks. Moreover,
the resultant stego-images demonstrate good visual quality in terms of Peak Signal-to-Noise
Ratio (PSNR). Nevertheless, the hiding capacity which is achieved is still limited due to the
fact that only part of the image serves to hide the embedding map.

Keywords: DCT, DWT, Embedding map, Image files, Steganography, SURF.

1. INTRODUCTION

In the modern era, computers and the internet are major communication media that
bring the different parts of the world together as a single global virtual world. As a result,
people can easily exchange information, while distance is no longer a barrier to
communication; however, the safety and security of long-distance communication remains an

22


issue. This is particularly important in the case of confidential data. The need to solve this
problem has led to the development of steganography techniques. Steganography is a
powerful (security) tool that provides a high level of security; particularly when it is
combined with cryptography [1]. Unlike cryptography, where the main goal is to secure
communications from an eavesdropper, steganographic techniques strive to hide the presence
of the message itself from an observer. Steganography does not replace cryptography; it
rather enhances security using its obscurity features.
Steganography is the art and science of concealing information in an appropriate
multimedia carrier, such as, image, audio and video files. It can be supposed that if a feature
is visible, the point of attack is evident. Therefore, the goal is to always hide the very
existence of the embedded data [2].
Steganography has many useful applications, e.g. in the copyright control of
materials, enhancing the robustness of image search engines and in smart IDs (identity cards)
where individuals’ details are embedded in their photographs. Steganography can be
characterized by three factors: undetectability (imperceptibility), robustness, and hiding
capacity [2]. It is not possible to maximize robustness, imperceptibility, and capacity
simultaneously; therefore, an acceptable balance of these items must be met by the
application. When steganography is used as a method for hiding communication,
imperceptibility becomes the most important requirement, while robustness and possibly
capacity can be sacrificed [4]. A number of ways exist to hide information in digital images.
Some of the common approaches include: Least Significant Bit insertion (LSB), Masking and
filtering, and Algorithms and transformations. Each of these techniques can be applied, with
varying degrees of success, to different image files [5]. For instance, LSB manipulation is a
quick and easy way to hide information; however, it is fragile to small changes resulting from
either the process of image processing or from the lossy compression. Masking techniques
embed information in significant areas so that the hidden message is more integral to the
cover image than being a mere hidden message in the “noise” level. Consequently, masking
techniques are more robust than LSB insertion with respect to compression, cropping, and to
some image processing. Hence, they are more suitable for use in digital watermarking [6].
Other more robust methods of hiding information in images include applications that
involve a manipulation of mathematical functions and image transformations. The widely
used transformational functions include DCT, Discreet Fourier Transform (DFT), and DWT
[7-12]. The basic approach to hiding information with DCT, DFT or DWT involves
transforming the cover image, tweaking the coefficients, and then inverting the
transformation. If the choice of coefficients is good and the size of the changes manageable,
then the result will be very close to the original [7].
Recently, Mali et al. proposed a robust DCT-based steganographic scheme via a
powerful coding framework that allows the dynamic choice of hiding locations and the
embedding of low and medium DCT coefficients [13]. The robustness of this scheme not
only comes from exploiting low and medium DCT coefficients for hiding data, but also
mainly from the redundancy, whereby the payload bits are repeated n times in order to add
robustness to the system. However, the scheme has a severe drawback that results in a loss of
information. In this paper, Mali et al.’s scheme and its drawbacks will be presented first.
Then, a proper modification is proposed to overcome the adopted scheme drawbacks and
make it more applicable using the embedding map.
This paper is organized as follows. In section 2 the related work is presented. In
section 3, our modification is proposed. The experimental results are presented in section 4.
Finally, the discussion and conclusion are given in sections 5 and 6 respectively.

23


2. RELATED WORK

DCT has been used widely for steganography and watermarking purposes. The DCT-
based methods hide data bits in significant areas of the cover-image in order to make them
more robust to attacks. Generally, DCT is applied to image blocks of 8×8 pixels, and selected
coefficients, of some selected blocks, are used to hide data bits. The coefficients are modified
differently in order to reflect an embedding of “0” or “1”.
Recently, Mali et al. [13] proposed a robust and secure method for embedding a high
volume of text information in digital cover-images without leaving perceptual distortion. It
has been found that this method is robust in combating intentional or unintentional attacks
such as image compression, tampering, resizing, filtering and AWGN. Fig. 1 shows the
steganographic data hiding system proposed by Mali et al.
Mali et al.’s scheme consists of two main stages: processing the data to be embedded
and embedding the data. In the first stage, the pure payload bits undergo three processes:
1- Encryption: to secure the data;
2- Redundancy addition: to reduce the bit error rate (BER); and
3- Interleaving: to ensure that the redundant bits are spread all over the image.
It is unnecessary to go through the details of this stage, and the reader can refer to
their algorithms for the details. Both redundancy and interleaving are responsible for the
recovery of the robust data at the receiver end. Nevertheless, overall robustness also depends
on the embedding procedures. For this reason, the embedding procedures and their drawbacks
are discussed in the following section.

Figure 1: General steganographic system proposed by Mali et al.

2.1 Mali’s Embedding Procedures

After processing the data to be embedded, the inputs to the embedding system are a
cover-image file (C), the processed text (FBS), Energy Threshold Factor (‫ ,) ⌃ ݓ‬and JPEG
quality factor (ܳ‫ .)ܨ‬The embedding phase can be summarized in the following section. The
reader can refer to the original paper for additional details.

24


Step 1: Divide the image into 8×8 non-overlapping blocks so that DCT is applied to each
block ܽ௜௝ to get ‫ܥ‬௜௝ as:

‫ܥ‬௜௝ = ‫ܶܥܦ‬൫ܽ௜௝ ൯ (1)

Where ݅, ݆ = ሼ0, 1, 2, … … . , 7ሽ

Step 2: Calculate the Energy of each block as:

଻ ଻
ଶ
‫ = ܧ‬෍ ෍ฮ‫ܥ‬௜௝ ฮ ∀݅, ݆ = ሼ0, 1, 2, … … … , 7 ሽ, (݅, ݆) ≠ 0 (2)
௜ୀଵ ௝ୀଵ

Step 3: Calculate the Mean Value of Energy (‫ )ܧܸܯ‬of the image using the
equation:

ଵ
‫ =ܧܸܯ‬஻ ∑஻ ‫ܧ‬௕
ଵ
(3)

where ‫ = ܤ‬Total number of blocks and ܾ = block number.

Step 4: Identify the Valid Blocks ܸ‫ ,ݏܤ‬which satisfy the Energy Threshold Criteria,
‫ , ்ܧ ≥ ܧ‬where ‫.ܧܸܯ × ⌃ ݓ = ்ܧ‬

Step 5: The coefficients of all VBs are quantized by dividing them according to their
respective elements of the quantization matrix as:

‫ܥ‬௜௝
‫ܥ‬௜௝ ⌃ = ொி ∀݅, ݆ = ሼ0, 1, … … , 7ሽ
(4)
‫ܯ‬௜௝

where, ‫ܥ‬௜௝ ⌃ is the quantized coefficient matrix, ‫ܯ‬௜௝ ொி is the ݆݅ ‫ݐ‬ℎ element of the
quantization matrix for a given value of QF.

Step 6: Identify the Valid DCT Coefficients (VCs), which satisfy the non-zero criteria
(‫ܥ‬௜௝ ≠ 0) and which fall into the lower and middle frequency band.

Step 7: The coefficients of all VCs are scanned in a zigzag fashion to get the one
dimensional vector ‫ܥ‬௞ . The process of embedding data will then be completed by
changing the quantized non-zero DCT coefficients, where the odd value for ‘bit =
0’ or the even value for ‘bit = 1’. The coefficients with the hidden bits ݀௞ are

25


given by,

ܱ݀݀ ‫ܥ‬௞ ⌃ , ݂݅ ܾ݅‫0 = ݐ‬ (5)
݀௞ ⌃ = ቊ
‫ܥ ݊݁ݒܧ‬௞ ⌃ , ݂݅ ܾ݅‫1 = ݐ‬

Step 8: The hidden coefficients ݀௞ ⌃ are reversely scanned to form an 8 × 8 matrix. It is
then multiplied by the JPEG quantization matrix to obtain unquantified
coefficients ‫ܥ‬௜௝ .
Step 9: Apply inverse DCT to each block, and reconstruct the image as stego-
image.

From the steps above, it is obvious that the extraction phase depends on identifying
the blocks that have been used for the correct embedding. Misidentifying those blocks will
cause a loss of portion of the embedded data or extracting unwanted data. During the
extracting phase, the blocks that have been used for embedding data should be identified first.
Such a process is achieved by the following two steps:
First: The energy of the block (‫ )ܧ‬should be ≥ ‫( ்ܧ‬step 4).
Second: The lower and middle DCT coefficients of the block should satisfy the non-zero
criteria (step 6). If the algorithm fails to identify the blocks in any of the steps, the embedded
data is extracted incorrectly.
The algorithm has been implemented and simulated through the use of different
images and randomly generated data. The results showed that the algorithm may misidentify
the blocks in some cases. For example, in one of the experiments on standard image ‘Lena’, it
was noted that the embedding process changed the ‫ ܧܸܯ‬and ‫ ܧ‬values. As a result, the
embedded block could not be identified during the extraction phase because ‫ . ்ܧ < ܧ‬In
another situation, reconstructing the stego-image involved a rounding operation to get the integer
pixel values. The rounding operation may turn some of non-zero coefficients to zero coefficients or
vice versa and also will cause misidentification of the blocks which carry the data; Fig. 2 illustrates
the above mentioned case.

71 68 75 85 75 75 67 81 71 70 83 85 79 80 58 83

71 67 75 80 73 74 77 88 66 61 75 73 70 77 70 92

77 78 75 79 81 80 90 100 73 68 72 69 73 81 85 103

78 74 75 69 82 92 100 112 81 66 76 65 76 94 97 111

80 75 85 81 88 99 109 121 85 65 89 84 84 104 109 116

80 81 84 88 100 112 122 133 83 65 86 92 96 119 125 127

77 81 95 105 113 121 135 141 86 66 97 109 105 126 140 134

81 91 105 114 122 131 143 144 100 82 111 119 112 135 148 135

‫ 076,565 = ܧܸܯ‬and ‫581,865 = ܧ‬ MVE = 566,780 and E = 560,244
‫்ܧ > ܧ → 1 = ⌃ ݓ‬ ‫்ܧ < ܧ → 1 = ⌃ ݓ‬
(a) (b)
Figure 2: This block is not detected because E <MVE: (a) the original block, (b) after
embedding.

26


Table 1 illustrates the number of misidentified blocks after applying the Mali et al.
algorithm on image ‘Lena’ using different quality factor values (ܳ‫ )ܨ‬and different energy
threshold values (‫ .) ⌃ ݓ‬Such a problem may affect the integrity of the embedded data;
especially, when one cannot identify which blocks have been misidentified. To solve the
problem of block misidentification, an embedding map (location map) is proposed in this
regard. The concept of the embedding map has been used in many data hiding techniques to
correctly identify the location of the blocks or regions where data has been embedded [14-
15]. That is to say, not the whole image is used for embedding. In the following section, Mali
et al.’s algorithm is modified by incorporating an embedding map.

Table 1: Misidentified blocks after applying Mali’s algorithm on image ‘Lena’

Number of misidentified blocks Percentage of misidentified blocks
QF = QF = QF = QF =
50% 75% QF = 100% 50% 75% QF = 100%
0.5 35 27 51 %1.17 %0.90 %1.71
0.6 30 28 47 %1.11 %1.03 %1.73

࢝⌃
0.7 35 27 39 %1.39 %1.07 %1.55
0.8 21 23 44 %0.90 %0.98 %1.88
0.9 42 14 39 %2.00 %0.67 %1.85
1 18 21 40 %0.98 %1.15 %2.19

3. THE PROPOSED ALGORITHM

The proposed algorithm is mainly based on that of Mali et al. In order to overcome
the problem of block misidentification, an embedding map technique is introduced to assure
the extraction of the embedded data correctly. Exploiting an embedding map implies
generating a binary map of a size equal to the number of blocks in the image. If the image
௠ ௡
size is ݉ × ݊, then, the embedding map size is ( × ), where the block size is 8 × 8
଼ ଼
pixles. Each block in the image is represented by a bit in the embedding map, and if the bit is
‘1’, this means that the corresponding block is used for embedding data and vice versa. The
embedding map will be concealed in specific regions of the image, while the data will be
concealed in different regions. The regions, i.e. the blocks, in which the data is hidden are
determined according to Mali et al.’s method. Although the embedding map can be
embedded in some predefined regions selected by the user, this option may affect the data
security because the same regions are used every time. Therefore, it is preferable to adopt a
more secure method to hide the embedding map, as described in the next section.

3.1 Hiding the Embedding Map

The embedding map is necessary to initiate the extracting phase, which means that it
must be concealed in the image in an extremely secure and robust way. To achieve these
objectives, two powerful techniques are used. The first technique is used to guarantee the
security of the embedding map and is performed by selecting the regions to be embedded in a
dynamic way, depending on the key-points of the image. For this purpose, Speed-Up Robust
Features (SURF) is used to extract the distinctive local features in the image and to produce
the key-point descriptors that demonstrate those features. Feature vectors/descriptors are

27


invariant to rotation, translation, and scaling; they are further partially invariant to
illumination changes and are robust to local geometric distortion [16]. Each feature vector has
some information that describes its corresponding key-point. The important part of the
information for the algorithm is the coordinates of the center of the key-point and its scale.
Twelve non-overlapping key-points with the highest scales are selected and used to hide the
embedding map. These key-points are the most robust features in the image and can be
detected even when the stego-image undergoes different types of operations, such as JPEG
compression and Gaussian noise. As to the second requirement, robustness, a DWT-based
embedding technique is adopted in which the data is embedded in a content-based manner.
For more details on using SURF and content-based embedding, the reader can refer to [17,
18].

3.2 Embedding Phase

Having introduced the embedding map, the proposed algorithm can be described
through the following steps:
Step 1: The SUR is applied to the image, and 12 key-points with the highest scales are
identified. The center coordinates of the key-point defines 12 square regions of
32 × 32 sized pixels with the same center coordinates. These regions are used to hide
the embedding map.
Step 2: The image is divided into 8×8 non-overlapping blocks so that DCT is applied to each
block ܽ௜௝ to get ‫ܥ‬௜௝ , as in Equation (1). Notice that the blocks that intersect with the 12
square regions that have been obtained in Step 1 are discarded as those regions are
used to hide the map not the data.
Step 3: The blocks are scanned, and the blocks which can be used for embedding are found
(according to Mali et al.’s algorithm), as described in section 2.1.
Step 4: An embedding map is built to indicate the blocks in which the data bits will be
embedded.
Step 5: The secret data is embedded in the blocks defined in Step 3 by modifying the DCT
coefficients according to Mali et al.’s algorithm.
Step 6: The embedding map is embedded in the 12 square regions defined in Step 1 by
modifying the DWT coefficients (in a content-based manner), as described in section
3.1. Notice that addition of the redundant bits and interleaving techniques have been
used to prepare both data and the embedding map to be hidden. Fig. 3 illustrates the
process of embedding data.

Figure 3: The process of embedding data using the proposed algorithm
28


4. EXPERIMENTAL RESULTS

To assess the performance of the proposed algorithm, experiments were performed on
three standard grayscale images; ‘Lena’, ‘Boat’, ‘and ‘Gold hill’, with the size 512 × 512, as
shown in Fig. 4. In order to evaluate the reliability of the proposed algorithm compared to
Mali et al.’s algorithm, different levels of attack were applied on the stego-image. The attacks
involved were JPEG compression and AWGN. A comparison between the proposed
algorithm and Mali et al.’s algorithm in terms of robustness is presented in Tables 2 and 3.

Table 2 A comparison between the proposed algorithm and Mali’s algorithm in terms of
reliability (Number of misidentified blocks). (QF = 75% & w ⌃ = 0.5)

Number of misidentified blocks
The proposed algorithm Mali's Algorithm
Gold Gold
Lena Boat hill Lena Boat hill
No attack 0 0 0 27 15 17
JPEG 100% 0 0 0 -14 -9 -5
JPEG 80% 0 0 0 316 249 228
Gaussian Noise 45dB 0 0 0 -26 -12 -5
Gaussian Noise 35dB 0 0 0 -32 -12 2

Table 3: A comparison between the proposed algorithm and Mali’s algorithm in terms of
reliability (Number of misidentified blocks). (QF = 75% & ‫)8.0 = ⌃ ݓ‬

Number of misidentified blocks
Gold Gold
No attack 0 0 0 23 14 32
JPEG 100% 0 0 0 -12 -3 6
JPEG 80% 0 0 0 250 247 175

29


It is worth mentioning that the number of missed blocks can be calculated by
subtracting the number of the detected blocks in the receiving end from the blocks that have
been used for secret data embedding (the actual embedding blocks). Accordingly, when the
number of the detected blocks is less than the actual number of the embedding blocks, the
result will be a positive number. On the other hand, due to the applied attacks to the stego-
image, erroneous blocks may be detected. As a result, the number of the detected blocks may
be more than the actual number of embedding blocks. This is the reason behind having
numbers with negative values representing the misidentified blocks, as shown in Tables 2 and
3.

In terms of imperceptibility, the visual quality of the obtained stego- image with the
proposed algorithm is better compared to those obtained by Mali et al.’s algorithm as
illustrated in Table 4. The visual quality is measured by the PSNR, as given in (6)

ெ஺௑಺ మ
ܴܲܵܰ(‫ܫ ,ܫ‬ௌ ) = 10 ݈‫݃݋‬ଵ଴ భ ೘షభ ∑೙షభฮூ(௜,௝)ି ூ మ (6)
∑ ೄ(೔,ೕ) ฮ
೘೙ ೔సబ ೕసబ

Where ‫ ܫ‬is the original image; ‫ܫ‬ௌ is the stego-image; ‫ܺܣܯ‬ூ is the maximum possible
pixel value of the image ‫ .ܫ‬The results obtained demonstrated that the proposed algorithm has
a better visual quality. For the purpose of evaluation, another comparison was made in terms
of the hiding capacity. The capacity is measured as the number of payload bits that can be
embedded in the image and retrieved successfully. The obtained results are shown in Table 5.

Table 4: A comparison between the proposed algorithm and Mali’s algorithm in
terms of PSNR

PSNR values (dB) @ QF = 75%
Gold Gold
0.5 37.28 34.10 36.42 33.19 32.66 33.36
0.6 37.62 34.24 36.98 33.56 32.78 33.82

࢝⌃
0.7 38.33 34.67 37.61 33.91 32.94 34.28
0.8 38.37 35.10 38.51 34.22 33.18 34.89
0.9 38.79 35.74 39.18 34.69 33.53 35.58
1 39.74 36.76 39.53 35.31 34.16 36.57

30


Table 5: A comparison between the proposed algorithm and Mali’s algorithm in terms of
hiding capacity

Hiding Capacity (bits)
Gold Gold
0.5 25,536 46,144 30,016 83,664 93,996 80,304
0.6 22,400 44,352 25,536 75,936 91,196 72,072

࢝⌃
0.7 19,264 38,976 21,056 70,476 87,892 64,344
0.8 18,816 33,152 16,128 65,688 83,272 56,196
0.9 15,680 25,088 12,096 58,884 76,580 47,796
1 11,648 16,128 10,752 51,212 66,724 38,444

‘Lena’ ‘Boat’ ‘Gold hill’

Figure 4: The images used for evaluation
5. DISCUSSION

In this paper, an interesting algorithm proposed by Mali et al. was reviewed. Their
algorithm shows adequate levels of robustness due to combining DCT and adding
redundancy bits. However, some of the blocks that carry data might be misidentified during
the extraction process. The present paper aims at enhancing the reliability of the original
algorithm by overcoming the problem of the misidentified blocks. To do so, an embedding
map has been adopted to indicate the location of the blocks, which have been used for
embedding. This means that some regions of the image will be exploited to hide data, while
others will be used to hide the embedding map. The blocks in which the data is concealed are
determined according to Mali et al.’s algorithm. The regions in which the embedding map is
concealed are determined in an extremely dynamic way to increase the security of the
algorithm. This goal has been achieved by the use of the SURF technique, which is used to

31


find the robust key-points of the image, due to the fact that each image has different key-
points. The embedding map is incorporated in those regions using a DWT-based method.

The experimental results in Tables 2 and 3 show that the proposed algorithm can
overcome the problem of block misidentification, even when the stego-image undergoes
JPEG compression or Gaussian noise. The high reliability of the algorithm in identifying the
blocks comes from the ability of SURF to detect the key-points, even after applying the
attacks. The DWT-based embedding technique also plays an important role in keeping the
embedding map intact. Nevertheless, strong attacks, such Gaussian noise may cause the
blocks to be slightly misidentified, as is the case in the image of ‘Boat’ in Table 4.

In comparison to Mali et al.’s algorithm, the proposed algorithm shows a better visual
quality in terms of PSNR, as shown in Table 4. This can be said for the three images used for
testing and with all (‫ ) ⌃ ݓ‬values used. However, Table 5 indicates that the hiding capacity of
the proposed scheme is somewhat lower than that achieved by Mali et al.’s algorithm. The
reason behind the capacity reduction is the exploitation of some regions of the image to hide
the embedding map.

The reason for the robustness of the algorithm in terms of Bit Error Rate (BER is not
tackled in this paper) is that the same embedding technique used in Mali et al.’s algorithm for
hiding the data has been used in the present study, which means that both algorithms have the
same level of robustness. The aim was also to enhance the reliability of Mali’s algorithm.
(Increasing the robustness is out of the scope of this paper).

6. CONCLUSIONS

The original algorithm presented by Mali et al. gives very good results in terms of
robustness; however, the algorithm cannot be considered reliable as some data may be lost. In
order to overcome the problem of lost data due to misidentified blocks, an embedding map
was used to specify the location of the blocks which were used for embedding secret
message, as the embedding map is very important to start the extracting phase correctly and
accurately. Any loss in the embedding map will in turn lead to data loss. Consequently, the
embedding map is hidden using SURF and DWT to assure robustness and to increase the
security level of the proposed system. The secret data can be embedded using the original
algorithm (DCT-based) proposed by Mali et al. While the experimental results show the
ability of the new algorithm to overcome the problem of lost data even with JPEG

32


compression or Gaussian noise, exploiting the embedding map reduces the available hiding
capacity. To increase this capacity, more embedding techniques with high hiding capacity
should be considered to hide the embedding map as this will permit more data (messages) to
be embedded in the image and retrieved effectively. Moreover, more possible attacks should
be investigated.
REFERENCES
[1] S.A.Halim and M.F.A.Sani, Embedding using spread spectrum image steganography with
GF (2m), Proc. IMT-GT-ICMSA, Kuala Lumpur, Malaysia, 2010, 659-666.
[2] A. Cheddad, Steganoflage: A new image steganography algorithm, doctoral diss.,
University of Ulster, Northern Ireland, UK, 2009.
[3] A. Cheddad, J. Condell, P.M. Kevitt, Digital image steganography: Survey and analysis
of current methods, Signal Processing Journal, 90(3), 2010, 727-752.
[4] B, Li, J. Huang, and Q.Y. Shi, A Survey on image steganography and steganalysis,
Journal of Information Hiding and Multimedia Signal Processing, 2(2), 2011, 123-138.
[5] N.F. Johnson and S. Jajodia, Exploring steganography: seeing the unseen, IEEE
Computer Journal, 31(2), 1998, 26-34.
[6] K. Curran and K. Baily, An evaluation of image based steganography methods,
International Journal of Digital Evidence, 2(2), 2003, 1-40.
[7] A. Nag, S. Biswas, D. Sarkar, and P.P. Sarkar, A novel technique for image
steganography based on Block-DCT and Huffman Encoding, International Journal of
Computer Science and Information Technology, 4(6), 2010, 103-112.
[8] C.C. Chang, T.S. Chen, and L.Z. Chung, A steganographic method based upon JPEG
and quantization table modification, Information Sciences Journal, 2002, 141(1,2), 123-138.
[9] M. Ashourian, R.C. Jain, and Y.H. Ho, Dithered quantization for image data hiding in
the DCT domain, Proc. of IST2003, 2003, 171-175.
[10] M. Iwata, K. Miyake, A. Shiozaki, Digital steganography utilizing features of JPEG
images, IEICE Trans. Fundamentals, E87-A, 2004, 929–936.
[11] C.C. Chang, C.C. Lin, C.S. Tseng, and W.L. Tai, Reversible hiding in DCT-based
compressed images, Information Sciences Journal, 177(13), 2007, 2768-2786.
[12] C.C. Lin, P.F. Shiu, High capacity data hiding scheme for DCT-based images. Journal
of Information Hiding and Multimedia Signal Processing, 1(3), 2010, 123-138.
[13] S.N. Mali, P.M. Patil, and R.M. Jalnekar, Robust and secured image-adaptive data
hiding, Digital Signal Processing, 22(2), 2010, 314–323.
[14] J. Tian, Reversible data embedding using a difference expansion, IEEE Transactions on
Circuits and Systems for Video Technology, 13(8), 2003, 890-896.
[15] A.M. Alattar, Reversible watermark using the difference expansion of a generalized
integer transforms, IEEE Transactions on Image Processing, 13(8), 2004, 1147-1156.
[16] H. Bay, From wide-baseline point and line correspondences to 3D, doctoral diss., Swiss
Federal Institute of Technology, Switzerland, 2006.
[17] N. Hamid, A. Yahya, R.B. Ahmad, and O.M. Al-Qershi, Characteristic region based
image steganography using Speeded-Up Robust Features technique. Proc. of IEEE
International Conference on Future Communication Networks (ICFCN 2012), Iraq, 2012,
141-146.
[18] L. Li, J. Qian, and J.S. Pan, Characteristic region based watermark embedding with RST
invariance and high capacity. AEU-International Journal of Electronics and Communications,
65(1), 2011, 435-442.

33

An improved robust and secured image steganographic scheme

More Related Content

What's hot (18)

Viewers also liked (7)

Similar to An improved robust and secured image steganographic scheme (20)

More from iaemedu (20)

An improved robust and secured image steganographic scheme