High capacity histogram shifting based reversible data hiding with data compression

Proceedings of the International Conference on Emerging Trends in Engineering and Management (ICETEM14)
30 – 31, December 2014, Ernakulam, India
180
HIGH CAPACITY HISTOGRAM SHIFTING BASED
REVERSIBLE DATA HIDING WITH DATA
COMPRESSION
Athira Ravi1
, Kavitha N Nair2
1, 2
University College of Engineering, Muttom, Kerala – 685587
ABSTRACT
Histogram shifting (HS) is a useful technique of reversible data hiding (RDH).With HS-based RDH, high
capacity and low distortion can be achieved efficiently. This paper revisits the HS techniques and presents a reversible
data hiding method accompanied with an encryption method so as to ensure the security of the host image which is in
high demand as well as the security of the message or data hidden in the host image respectively. An image block
division technique is introduced to improve the embedding capacity and visual quality of the host image. To increase the
embedding capacity further, data compression technique is used in conjunction with the histogram shifting method. Two
lossless compression techniques, Huffman coding and LZW coding are used to compress the secret data.
Keywords: Embedding Performance, Histogram Shifting (HS), Reversible Data Hiding (RDH).
1. INTRODUCTION
Data hiding is a technique used to put a secret data in a host media (like images) with small changes in host. In
most of the data hiding schemes the cover image becomes distorted due to data hiding process and it cannot be retrieved
back to the original form. Thus the cover media is permanently distorted due to the data embedding. In some
applications, such as medical image processing and military image processing[1], retrieval of the original cover image
without any damage is a must, since these images have too process further. The process of retrieving the cover or host
image without any damage after the secret data extraction is known as Reversible data hiding.
Most of the proposed data hiding schemes are not reversible. Reversible data hiding can be done in many ways
like, Integer-to-Integer Wavelet Transform [2], Difference expansion [3], and Histogram modification [4].
Tian’s DE algorithm is an efficient work of RDH. In DE algorithm, the host image is divided into pixel pairs,
and the difference value of two pixels in a pair is expanded to carry one data bit. The original content restoration
information, a message authentication code, and additional data (which could be any data, such as date/time information,
auxiliary data, etc.) will all be embedded into the difference values. By exploring the redundancy in the image,
reversibility is achieved. This method can be applied to digital audio and video and can provide an embedding rate (ER)
up to 0.5 bits per pixel (BPP). The major drawback of Tian’s scheme is the lack of capacity control.
Ni et al. proposed a reversible data hiding method [5] based on histogram modification. In this method, each
pixel value is modified at most by 1, and thus the visual quality of marked image is guaranteed. This algorithm utilizes
the zero or the minimum points of the histogram of an image and slightly modifies the pixel gray scale values to embed
data into the image. It can embed more data than many of the existing reversible data hiding algorithms. The peak signal-
to-noise ratio (PSNR) of the marked image generated by this method versus the original image is guaranteed to be above
48 db. This lower bound of PSNR is much higher than that of many of the proposed reversible data hiding techniques.
INTERNATIONAL JOURNAL OF ELECTRONICS AND
COMMUNICATION ENGINEERING & TECHNOLOGY (IJECET)
ISSN 0976 – 6464(Print)
ISSN 0976 – 6472(Online)
Volume 5, Issue 12, December (2014), pp. 180-191
© IAEME: http://www.iaeme.com/IJECET.asp
Journal Impact Factor (2014): 7.2836 (Calculated by GISI)
www.jifactor.com
IJECET
© I A E M E

181
The main goal of this work is to implement a Histogram shifting (HS) based Reversible Data Hiding(RDH)
method that can provide a high embedding capacity with lowest distortion. First in section II the data embedding and
extraction process is explained. Also the image block division technique for improving marked image quality and the
data compression technique for improving embedding capacity is given. Results and discussions are given in section III.
Finally, concluding remarks are given in the last section.
2. METHODOLOGY
The proposed method presents a reversible data hiding method accompanied with an image block division
technique and data compression method so as to further increase the embedding capacity. An image encryption technique
is adopted to ensure the security of the host image and the secret data. For reversible data hiding, an efficient extension of
the histogram modification technique by considering the differences between adjacent pixels instead of simple pixel
value is used. A binary tree structure is used to solve the issue of communication of multiple peak points. To prevent
overflow and underflow, a histogram shifting technique that narrows the histogram from both sides adopted. To further
ensure the security of the host image the host image is encrypted using an encrypting algorithm that ensures reversibility.
2.1. Data Hiding Method
In the proposed method, for reversible data hiding, an efficient extension of the histogram modification
technique by considering the differences between adjacent pixels instead of simple pixel value is used. Since image
neighbour pixels are strongly correlated, the distribution of pixel difference has a prominent maximum. Hence there will
be lot of candidates for data embedding as shown in Fig.2. For the original histogram the count of the maximum pixel is
between 1400 and 1600. But for the difference image histogram the count is 14000.
Fig.2: a) Original histogram b) Shifted histogram
Images having an equal histogram, the histogram modification technique does not work well. While multiple
pairs of peak and minimum points are used for embedding, the pure payload is still a little low. Moreover, the histogram
modification technique carries with it an unsolved issue in that multiple pairs of peak and minimum points must be
transmitted to the recipient via a side channel to ensure successful restoration. In RDH schemes, large hiding capacities
can be obtained by repeated data hiding process. But the recipients are not able to retrieve both the embedded message
and the original image without the knowledge of peak points of every hiding process. By supplying a side
communication channel for the peak points this issue can be solved. But this side communication channel may extend the
embedded message length and therefore it may reduce the embedding capacity. So binary tree structure is introduced to
solve the issue of communication of multiple peak points.
Figure below shows an auxiliary binary tree for solving the issue of communication of multiple peak points.
Each element denotes a peak point. Assume that the number of peak points used to embed messages is 2௅
, where L is the
level of the binary tree. Once a pixel difference݀௜that satisfies ݀௜<2௅
is encountered, if the message bit to be embedded is
0, the left child of the node݀௜is visited. Otherwise, the right child of the node݀௜is visited. Higher payloads require the use
of higher tree levels, thus quickly increasing the distortion in the image beyond acceptable levels. However, the entire
recipient needs to share with the sender is the tree level L, because an auxiliary binary tree is proposed that predetermines
multiple peak points used to embed messages.

Proceedings of the International Conference on Emerging
Modification of a pixel may not be allowed if the pixel is saturated (0 or
underflow, a histogram shifting technique that narrows the histogram from both sides, as shown in figure below is
adopted. Assume that the number of peak points used to embed messages is
binary tree structure. Thus the histogram is shifted from both sides by
the pixel ‫ݔ‬௜that satisfies ݀௜ ≤ 2௅
shift by
After narrowing the histogram to the range
overhead bookkeeping information. For this purpose a one bit map as the location map is created, which is equal in size
to the host image. For a pixel having grayscale value in the
Otherwise, assign 1. The location map is loss
large increase in compression ability since pixels out of the range
be embedded into the host image together with the embedded message.
Fig.3
2.2.Embedding process
The embedding process involves several steps. For an N
value‫ݔ‬௜where‫ݔ‬௜denotes the grayscale value of the pixel, 0
1) Read the host image. Determine the level L of the binary tree.
2) Shift the histogram of the host image from both sides
overhead bookkeeping information that will be embedded into the image itself with payload.
3) Scan the image host image in an inverse s
݀݅ ൌ ൜
‫ݔ‬௜ ݂݅
‫ݔ׀‬௜ିଵ െ ‫ݔ‬௜‫ ,׀‬
International Conference on Emerging Trends in Engineering and Management (ICETEM14)
182
Fig.3: Auxiliary binary tree
Modification of a pixel may not be allowed if the pixel is saturated (0 or 255). To prevent overflow and
adopted. Assume that the number of peak points used to embed messages is 2௅
, where L is the level of the propo
binary tree structure. Thus the histogram is shifted from both sides by 2௅
units to prevent overflow and underflow since
2௅
units afterembedding takes place.
After narrowing the histogram to the rangeሾ 2௅
, 255-2௅
], the histogram shifting information is recorded as the
to the host image. For a pixel having grayscale value in the rangeሾ 2௅
, 255-2௅
], assign a value 0 in the location map.
The location map is loss lessly compressed by the run-length coding algorithm, which will yield a
large increase in compression ability since pixels out of the rangeሾ 2௅
, 255-2௅
], are few. The overhead information will
be embedded into the host image together with the embedded message.
Fig.3: a) Original histogram b) Shifted histogram
The embedding process involves several steps. For an N-pixel 8-bit grayscale host image H with a pixel
denotes the grayscale value of the pixel, 0 ≤ i ≤ N -1,‫ݔ‬௜ϵ [0, 255].
1) Read the host image. Determine the level L of the binary tree.
2) Shift the histogram of the host image from both sides by 2௅
units. The histogram shifting information is recorded as
overhead bookkeeping information that will be embedded into the image itself with payload.
3) Scan the image host image in an inverse s-order. Calculate the pixel difference ݀௜ between pixels
݂݅ ݅ ൌ 0 ,
‫ .݁ݏ݅ݓݎ݄݁ݐ݋‬
Trends in Engineering and Management (ICETEM14)
31, December 2014, Ernakulam, India
255). To prevent overflow and
, where L is the level of the proposed
units to prevent overflow and underflow since
the histogram shifting information is recorded as the
assign a value 0 in the location map.
length coding algorithm, which will yield a
are few. The overhead information will
bit grayscale host image H with a pixel
units. The histogram shifting information is recorded as
between pixels ‫ݔ‬௜ିଵand ‫ݔ‬௜.
(1)

4) Create location map using difference image same size as that of
image).
‫݌ܽ݉_݊݋݅ݐܽܿ݋ܮ‬ ൌ ൜
0,
1,
5) Compress the location map using run length encoding.
6) Convert the compressed location map to binary.
7) Read the message to be hidden and convert to binary.
8) Combine the message and location map in binary form.
9) Embed the combination of message and location map into histogram shifted image using pixel difference image as
follows.
Scan the whole image in the same inverse s
݀௜ ൒ 2௅
,shift ‫ݔ‬௜by 2௅
units.
‫ݕ‬௜ ൌ ቐ
‫ݔ‬௜ ݂݅ ݅ ൌ
‫ݔ‬௜ ൅ 2௅
, ݂݅ ݀௜ ൒ 2௅
ܽ݊݀ ‫ݔ‬௜ ൒ ‫ݔ‬
‫ݔ‬௜ െ 2௅
, ݂݅ ݀௜ ൒ 2௅
ܽ݊݀ ‫ݔ‬௜ ൏ ‫ݔ‬
Where ‫ݕ‬௜ is the watermarked value of pixel.
10) If݀௜ ൏ 2௅
, modify xi according to the message bit.
‫ݕ‬௜ ൌ ൜
‫ݔ‬௜ ൅ ሺ݀௜ ൅ ܾሻ, ݂݅
‫ݔ‬௜ିሺ݀௜ ൅ ܾሻ, ݂݅
Where b is a message bit to be embedded and b
After hiding the secret data using the embedding technique, the host image together with the hidden data is
encrypted as a whole to get the output image. At the receiver side, the encrypted image is read as the
decrypting the image the secret data can be extracted using the following extraction procedure.
183
4) Create location map using difference image same size as that of difference image (which is same size as that of host
݂݅ ‫ ݈݁ݔ݅݌‬ ∈ ሾ2௅
, 255 െ 2௅
ሿ
ܱ‫݁ݏ݅ݓݎ݄݁ݐ‬
5) Compress the location map using run length encoding.
6) Convert the compressed location map to binary.
7) Read the message to be hidden and convert to binary.
8) Combine the message and location map in binary form.
e and location map into histogram shifted image using pixel difference image as
Scan the whole image in the same inverse s-order. If
ൌ 0
‫ݔ‬௜ିଵ
‫ݔ‬௜ିଵ
is the watermarked value of pixel.
according to the message bit.
݂݅ ‫ݔ‬௜ ൒ ‫ݔ‬௜ିଵ
‫ݔ‬௜ ൏ ‫ݔ‬௜ିଵ
message bit to be embedded and b ϵ {0, 1}.
encrypted as a whole to get the output image. At the receiver side, the encrypted image is read as the
Fig.4: Flow diagram for data embedding
difference image (which is same size as that of host
(2)
e and location map into histogram shifted image using pixel difference image as
(3)
(4)
encrypted as a whole to get the output image. At the receiver side, the encrypted image is read as the input image. After

2.3. Extraction process
This process extracts both overhead information and payload from the
recovers the host image. Let L be the level of the proposed binary tree. For an N
pixel value‫ݕ‬௜, where ‫ݕ‬௜denotes the gray scale value of the
1) Scan the watermarked image W in an inverse s
2) If |‫ݕ‬௜− ‫ݔ‬௜ିଵ| < 2௅ାଵ
, extract message bit b by

Where ‫ݔ‬௜ିଵ denotes the restored value of
3) Restore the original value of host pixel
4) Repeat Step 2 until the embedded message is
5) Extract the overhead information from the extracted message. If a value 1 is assigned in the location i, restore to its
original state by shifting it by units. Otherwise, no shifting is required.
Fig.5. shows the complete flow diagram for d
2.4. Encryption and Decryption
To further ensure the security of the host image the host image is encrypted using an encrypting algorithm that
ensures reversibility. A key Based Algorithm using logistic map
sequence is generated using a logistic mapping.
key is used to encrypt and decrypt the image. T
each pixel in the image. The encryption scheme based on logistic Map has higher decorrelating ability.
‫ݔ‬ሺ݊ሻ ൌ 1 െ 2 ൈ ‫ݔ‬ሺ݊ െ 1ሻ ൈ
| |
| |
2 , if | | 2 and
2 , if | | 2 and
i i
i i i i i
i i
i i i i i
i
L L
i i i i i
L L
i i i
y x
y y x x
y x
y y x x
x
y y x y x
y y x
 
+ − < < 
 
 
− − < > =  
+ − ≥ <
− − ≥
,iy











184
This process extracts both overhead information and payload from the watermarked image and losslessly
recovers the host image. Let L be the level of the proposed binary tree. For an N-pixel 8-bit watermarked image W with a
denotes the gray scale value of the݅௧௛
pixel,0 ≤ i ≤ N-1, ‫ݕ‬௜ ϵ [0, 255].
1) Scan the watermarked image W in an inverse s-order.
, extract message bit b by
ܾ ൌ ቊ
0 ݂݅ ‫ݕ׀‬௜ െ ‫ݔ‬௜ିଵ‫ ݊݁ݒ݁ ݏ݅׀‬
1 ݂݅ ‫ݕ׀‬௜ െ ‫ݔ‬௜ିଵ ‫ ݀݀݋ ݏ݅ ׀‬
denotes the restored value of ‫ݕ‬௜ିଵ
3) Restore the original value of host pixel ‫ݔ‬௜by
4) Repeat Step 2 until the embedded message is extracted.
original state by shifting it by units. Otherwise, no shifting is required.
shows the complete flow diagram for data extraction process.
ey Based Algorithm using logistic map is used for encryption and decryption [6][7]. A key
sequence is generated using a logistic mapping. Image pixels are rearranged and XORed with the selected key.
key is used to encrypt and decrypt the image. The given difference equations is used to generate an 8
ൈ ‫ݔ‬ሺ݊ െ 1ሻ
Fig.5: Flow diagram for data extraction
11
1 1
11
1 1
1
1 1
1
1
| |
, if | | 2 andy
2
| |
, if | | 2 andy
2
2 , if | | 2 and
2 , if | | 2 and
Li i
i i i i i
Li i
i i i i i
L L
i i i i i
L L
i i i
y x
y y x x
y x
y y x x
y y x y x
y y x
+−
− −
+−
− −
+
− −
+
−
− 
+ − < < 
 
− 
− − < > 
 
+ − ≥ <
− − ≥ 1
, otherwise
i iy x−>
watermarked image and losslessly
bit watermarked image W with a
(5)
(6)
is used for encryption and decryption [6][7]. A key
Image pixels are rearranged and XORed with the selected key. The same
to generate an 8-bit binary "key" for
(7)

185
2.5. Block Division for Improving Marked Image Quality
To enhance the data hiding capacity and visual quality a block division technique is proposed. In the proposed
approach, the input image is divided into blocks and then histogram shifting is done on each block. Amount of
information that can be embedded within image blocks are more as compared with embedding within a single image.
This technique consists of three main stages. 1) Dividing the image into two blocks 2) Processing stage
3) Embedding stage. First stage consists of dividing the image into two main blocks. Processing stage includes
generating the histogram of each block and taking the difference of histogram after histogram modification. After
histogram modification, secret data embedding and extraction can be performed with the same embedding and extraction
algorithm which discussed earlier. In the previous method embedding and extraction is done with a single image. In
block division method data embedding is done after dividing the image into blocks. Data embedding and extraction is
performed with the two blocks separately.
There are so many advantages while considering the histogram of image blocks than a single image. It is
possible to distribute the embedded bits along the whole image. Image blocks have narrower histogram and thus it helps
in selecting the suitable peak and zero points which may increase the quality of watermarked image.
2.6. Data Compression for Improving Embedding Capacity
In the proposed technique, if the data embedded in the image is increased, the image quality deteriorates. So, we
cannot embed sufficiently large data into the cover image. To overcome this problem prior to embedding secret data is
pre-processed first and then this pre-processed data is embedded into the host image. For pre-processing data
compression techniques can be used [8].
Data compression involves encoding information using fewer bits than the original representation. The general
principle of data compression algorithms on text files is to transform a string of characters into a new string which
contains the same information but with new length as small as possible. In this thesis two lossless compression methods,
Huffman coding and LZW coding are used to compress the text data. Lossless algorithms are typically used for text, and
lossy for images and sound where a little bit of loss in resolution is often undetectable, or at least acceptable. The
compression efficiency of the two methods is compared with respect to data embedding capacity limit.
2.6.1. Huffman Coding
Huffman algorithm is the oldest and most widespread technique for data compression. It was developed by
David A. Huffman and used in compression of many type of data such as text, image, audio, and video. It is based on
building a full binary tree for the different symbols that are in the original file after calculating the probability for each
symbol and put them in descending order. After that, we derive the code words for each symbol from the binary tree,
giving short code words for symbols with large probabilities and longer code words for symbols with small probabilities.
Suppose that we have a test file that uses only five characters A, B, C, D, E. Frequency of each character is shown in the
table.
TABLE I: Descending frequencies for symbols
Symbol Frequency
E 32
D 27
C 12
B 12
A 17
Each character is considered as a node. Start by choosing the two smallest nodes, combine them into a new tree
and the root of this new tree is the sum of the weight of the small nodes. Replace those two nodes with the new tree. By
repeating this, the complete Huffman tree can be obtained as shown in Fig.6. Suppose that we have a test file that uses
only five characters A, B, C, D, E. Frequency of each character is shown in the table I.

186
Fig.6: Huffman tree
Now we assign codes to the tree by placing a 0 on every left branch and a 1 on every right branch. A traversal of
the tree from root to leaf gives the Huffman code for that particular leaf character. Code word is only completed when
leaf node is reached. Then we get the code word for each symbol from the binary tree as in Table II.Note that no code is
the prefix of another code. With this Huffman code the given Text: EAEBCD can be coded as 11001101001110. Since
there are six characters in the text when ASCII encoding is used text is 48 bit long. But with Huffman coding the same
text require only 14 bits. Thus Huffman encoding can be effectively used to compress data. Due to the prefix property of
the Huffman code, the codes are uniquely decodable.
TABLE II: Code words for each symbol
2.6.2. LZW Coding
LZW is a general compression algorithm capable of working on almost any type of data. LZW compression
creates a table of strings commonly occurring in the data being compressed, and replaces the actual data with references
into the table. The table is formed during compression at the same time at which the data is encoded and during
decompression at the same time as the data is decoded. The algorithm is surprisingly simple. LZW compression replaces
strings of characters with single codes. It does not do any analysis of the incoming text. Instead, it just adds every new
string of characters it sees to a table of strings. Compression occurs when a single code is output instead of a string of
characters. During encoding, LZW algorithm identifies repeated sequences in the data and replaces them with a unique
code in the dictionary as shown in Fig.3.10. Data compression occurs when all characters except the last character is
replaced with the index found in dictionary. During decompression the index is replaced by the corresponding entry in
the dictionary.
2.6.3. Lower Bound of PSNR
The pixel x୧whose differenced୧is larger than peak point will be either increased or decreased by 1 in the data
embedding process with one peak point. Therefore, in the worst case, all pixel values will be increased or decreased by 1.
That is, the resulted the mean squared error (MSE) is (N-1)/N, which is almost equal to 1 when N is large enough. Thus,
the lower bound of PSNR for the watermarked image generated from the embedding process with one peak point is
ܴܲܵܰሺ݀‫ܤ‬ሻ ൌ 10 × ݈‫݃݋‬ଵ଴ ቀ
ଶହହమ
ெௌா
ቁ ≥ 48.13 dB (8)
Symbol Frequency Code word
E 32 11
D 27 10
A 17 0
B 12 10
C 12 11

187
As a result, the lower bound of PSNR for the watermarked image generated by our proposed algorithm with one
peak point is theoretically proved larger than 48 dB, which is also supported by numerous experiments. The MSE and
PSNR are better will be the quality of the watermarked or reconstructed image. Greater the value of the peak point i.e.
smooth regions, more number of bits can be embedded within the image.
3. RESULT AND DISCUSSION
Performance of the proposed algorithm is tested with six different datasets of size 256×256 with 8 bit resolution.
The method is applied on six test images of size 256×256 as shown in Fig.7.
Fig.7: Test images
Variation of the PSNR for different values of L (0 to 4) is analyzed. Table 4.1 summarizes the variation of
PSNR (dB) with tree level from 0 to 4 for different images. As table shows, distortion of image increases and PSNR
values decreases with rise in the value of L.
The output images obtained upon the application of proposed method on image 1 for L value equals 1 is
TABLE III: Variation of PSNR (dB) for different values of L
given below. Obviously, the watermarked image hardly can be distinguished from the original image. The host image
can be reconstructed without any damage.
Host Image
256*256
PSNR values for different L values
0 1 2 3 4
Image 1 52.04 47 42.52 38.38 34.72
Image 2 52.13 46.84 42.10 38.14 34.34
Image 3 51.88 45.91 40.79 36.48 34.05
Image 4 54.69 50.54 46.63 43.37 41.04
Image 5 51.50 45.91 40.79 36.48 34.05
Image 6 51.77 46.44 41.73 37.61 34.18
Image 5 Image 6Image 4
Image 1 Image 2 Image 3

188
Fig.8: Output images a) Input image b) Embedded image c) Encrypted image d) Decrypted image
e) Reconstructed image
To improve the visual image quality of the watermarked image a block division technique is adopted. Here data
embedding is done after dividing the image into two blocks. Firstly, image is divided into two blocks as shown in Fig.9.
Then histogram of each block is plotted. After histogram modification, data is embedded into each block using the
proposed embedding algorithm.
Fig.10. shows the histogram of image after dividing it into two blocks and histogram of them after histogram
modification. Histogram of individual image blocks makes it possible to distribute the message bits along the whole
image and also improves the image quality.
Fig.9: Image after block division
Table IV shows that PSNR is more when embedding is performed after dividing the image into blocks when
compared with the embedding performed in a single image. Higher the PSNR, higher will be the image quality. Thus
block division technique can effectively use to improve the marked image quality.
(a) (b)
(e)(d)
(c)

189
Fig.10: Histogram of the input image and image blocks
TABLE IV: Variation of PSNR for a single image and image after block division
Tree Level
L
PSNR of whole
Image
PSNR of two blocks
Average PSNR after
block division
0 56.656
57.082
56.683
56.284
1 52.113
52.650
52.116
51.583
2 47.553
48.229
47.586
46.942
3 43.288
44.079
43.344
42.609
4 39.664 40.475 39.695
Fig.11. shows the data embedding and extraction process for block division technique. Since the secret data is
embedded into the two image blocks separately, there will be two embedded image blocks. The two reconstructed image
blocks after secret data extraction can combine without any distortion and the complete reconstructed image can be
obtained as shown in figure.

190
Fig.11: Output images for block division method a) Input image b) Embedded image for left block c) Embedded image
for right block d) reconstructed image for left block e) Reconstructed image for right half
f) Complete reconstructed image
The histogram shifting technique based on pixel differences itself can provide a higher embedding capacity. To
further improve the data hiding capacity secret data is compressed before data embedding. Two lossless data
compression methods Huffman encoding and LZW coding are used.
Table V shows the variation of PSNR and MSE with and without Huffman coding. The PSNR values with
Huffman data compression are greater than that without compression. Thus data compression using Huffman coding can
provide higher embedding capacity.
Similar results can be obtained with the LZW coding.PSNR values with LZW coding is higher than that without
compression.
TABLE V: Variation of PSNR and MSE with and without Huffman coding
L With data compression
(Huffman Coding)
Without data compression
PSNR MSE PSNR MSE
0 52.040 0.04 52.040 0.04
1 47.094 1.267 47.039 1.285
2 42.534 3.627 42.511 3.647
3 38.392 9.414 38.377 9.447
Table VI compare the embedding capacity of the two compression methods.Higher PSNR values can be
obtained with the Huffman coding compared with LZW coding. Thus better compression is occurring with Huffman
coding. Also Huffman coding is easier to implement.
TABLE VI: PSNR value comparison for Huffman coding and LZW coding
L
Without data
compression
Huffman
coding
LZW coding
0 52.048 52.048 52.048
1 47.135 7.139 47.140
2 42.559 42.562 42.560
3 38.408 38.409 38.407
Fig.12 shows the comparison of tree level, L versus the peak signal to noise ratio for the text data without data
compression, with Huffman coding and LZW coding. PSNR values are plotted against tree level, L.Higher PSNR values
with data compression schemes indicates that, when the text data is
(a) (b) (c)
(d) (e) (f)

(a)
Fig.12: Comparison of PSNR values a) With Hu
compressed more data bits can be embedded with
the data compression techniques together with the histogram shifting techniques can improve the embedding capacity
and watermarked image quality.
4. CONCLUSION
The proposed method presents a reversible data hiding method accompanied with an encryption method so as to
ensure the security of the host image and the security of the message or data hidden in the host image. For reversible data
hiding, an efficient extension of the histog
pixels instead of simple pixel value is used. A binary tree structure is used to solve the problem of communicating pairs
of peak points. Distribution of pixel differences is used t
A histogram shifting technique is used to prevent overflow and underflow. The method ensures reversibility by showing
higher rate of PSNR values. The encryption method ensures reversibility
block division technique helps to distribute the message bits along the whole image and improves the visual quality of
the image and the hiding capacity. Two lossless data compression techniques, Huffman enc
used in conjunction with HS to further improve the embedding capacity.
In the future, the research can be extended in the following direction. The one are to promote data capacity and
stego-image quality at the same time. The propo
image, in the future the wasting capacity of extra information can reduce.
REFERENCES
[1] R.norcen, M.podesser, A.pommer,
Image Data”, Computers in Biology and Medicine 33,
[2] Sunil Lee, Chang D. Yoo, “Reversible Image Watermarking Based on Integer
2007.
[3] J. Tian, “Reversible data embedding using a difference expansion,” I
vol. 13, no. 8, pp. 890–896, Aug.2003.
[4] ZhenfeiZhaoa, HaoLuoc,, Jeng
modification and sequential recovery “International Journal of Electronics and
[5] Z. Ni, Y. Q. Shi, N. Ansari, and W. Su, “Reversible data hiding,”
vol. 16, no. 3, pp. 354–362, Mar.
[6] N.K. Pareek, VinodPatidar, K.K. Sud, Discrete chaotic cryptography using external
(2003) 75–82.
[7] G. Chen, Y. Mao, C.K. Chui, “A symmetric image encryption based on 3D chaotic maps”,
Fractals 21 (2004)749–761.
[8] Mikhail.J.Atallah, “Text data compression,” in Algorithms and Theory of computation
pp. 12-1 - 12-21, 2009.
[9] Shamim Ahmed Laskar and Kattamanchi Hemachandran, “Steganography Based
Efficient Data Hiding”, International
Issue 2, 2013, pp. 31 - 44, ISSN Print: 0976
191
(a) (b)
Comparison of PSNR values a) With Huffman coding b) With LZW coding
compressed more data bits can be embedded with the same image and hence it will improve the embedding capacity. So
presents a reversible data hiding method accompanied with an encryption method so as to
hiding, an efficient extension of the histogram modification technique by considering the differences between adjacent
of peak points. Distribution of pixel differences is used to achieve large hiding capacity while keeping the distortion low.
higher rate of PSNR values. The encryption method ensures reversibility of the host image in addition to security. Image
the image and the hiding capacity. Two lossless data compression techniques, Huffman enc
used in conjunction with HS to further improve the embedding capacity.
image quality at the same time. The proposed scheme still need to record extra information for restoring the cover
image, in the future the wasting capacity of extra information can reduce.
A.pommer, H.Schmidt and A.Uhl, “Confidential storage and Transmission of
Computers in Biology and Medicine 33, pp.277-292, 2003.
Reversible Image Watermarking Based on Integer-to-Integer Wavelet Transform
J. Tian, “Reversible data embedding using a difference expansion,” IEEE Trans. Circuits Syst. Video Technol.,
896, Aug.2003.
HaoLuoc,, Jeng-ShyangPand, “Reversible data hiding based on multilevel
recovery “International Journal of Electronics and Communications,
Z. Ni, Y. Q. Shi, N. Ansari, and W. Su, “Reversible data hiding,” IEEE Trans. Circuits Syst. Video Technol.
362, Mar. 2006.
N.K. Pareek, VinodPatidar, K.K. Sud, Discrete chaotic cryptography using external
G. Chen, Y. Mao, C.K. Chui, “A symmetric image encryption based on 3D chaotic maps”,
Mikhail.J.Atallah, “Text data compression,” in Algorithms and Theory of computation
nd Kattamanchi Hemachandran, “Steganography Based on Random Pixel Selection
International Journal of Computer Engineering & Technology (IJCET), Volume
44, ISSN Print: 0976 – 6367, ISSN Online: 0976 – 6375.
ffman coding b) With LZW coding
the same image and hence it will improve the embedding capacity. So
presents a reversible data hiding method accompanied with an encryption method so as to
ram modification technique by considering the differences between adjacent
o achieve large hiding capacity while keeping the distortion low.
of the host image in addition to security. Image
the image and the hiding capacity. Two lossless data compression techniques, Huffman encoding and LZW coding is
sed scheme still need to record extra information for restoring the cover
Confidential storage and Transmission of Medical
Integer Wavelet Transform”,
Trans. Circuits Syst. Video Technol.,
based on multilevel histogram
Communications, 2010.
Trans. Circuits Syst. Video Technol.,
N.K. Pareek, VinodPatidar, K.K. Sud, Discrete chaotic cryptography using external key, Phys. Lett. A 309
G. Chen, Y. Mao, C.K. Chui, “A symmetric image encryption based on 3D chaotic maps”, Chaos Solitons
Mikhail.J.Atallah, “Text data compression,” in Algorithms and Theory of computation Handbook., CRC press,
n Random Pixel Selection for
ournal of Computer Engineering & Technology (IJCET), Volume 4,

High capacity histogram shifting based reversible data hiding with data compression

Recommended

More Related Content

What's hot (20)

Similar to High capacity histogram shifting based reversible data hiding with data compression (20)

More from IAEME Publication (20)

Recently uploaded (20)

High capacity histogram shifting based reversible data hiding with data compression