compressed and highly secured reversible image data hiding ... · highly secured reversible image...

2749

www.ijifr.com Copyright © IJIFR 2015

Research Paper

International Journal of Informative & Futuristic Research ISSN (Online): 2347-1697

Volume 2 Issue 8 April 2015

Abstract

This system is proposed for digital images, mainly focussed on grey scale images. Usual algorithms just keep the PSNR values high to enhance the contrast of an image, where this system enhances the image quality as well as its security and decreases the bandwidth consumed. For embedding data in the image highest two peaks in the histogram are selected and then repeated the same process for performing histogram equalization. The data is embedded in the image, compressed and encrypted using chaotic encryption such that the data and image are completely recoverable. The system with the proposed algorithm was performed on two sets of images to prove its efficiency. It is found that, by embedding considerable amount of data into the image, the contrast of the image is being enhanced. It is proved that, this system works better than the three inbuilt MATLAB functions for contrast enhancement which as an add-on provides compression and encryption for better security.

Compressed And Highly Secured

Reversible Image Data Hiding With

Contrast Enhancement Paper ID IJIFR/ V2/ E8/ 081 Page No. 2749-2760 Research Area

Computer Sci.

& Engineering

Key Words Contrast Enhancement, Reversible Data Hiding, Histogram Modification,

Chaotic Encryption, Huffman Coding

Anju Mariyam Zacharia 1

M.Tech. Scholar

Department of Computer Engineering

Musaliar College of Engineering and Technology, Pathanamthitta-Kerala

Shyjila P.A. 2

Associate Professor


Musaliar College of Engineering and Technology,

Pathanamthitta-Kerala

Karthika J S 3

M.Tech. Scholar


Musaliar College of Engineering and Technology,

Pathanamthitta-Kerala

2750

ISSN (Online): 2347-1697 International Journal of Informative & Futuristic Research (IJIFR)

Volume - 2, Issue - 8, April 2015 20th Edition, Page No: 2749- 2760

Anju Mariyam Zacharia , Shyjila P.A., Karthika J S:: Compressed And Highly Secured Reversible Image Data Hiding With Contrast Enhancement

1. Introduction

Reversible Data Hiding (RDH) also referred to as Lossless data hiding, or invertible data

hiding is used widely in the field of signal processing. RDH is usually used to hide a piece

of data into an image to produce a marked image. The highlight of RDH is that it gives back

the original image after the data is being extracted from the marked image. This manner of

RDH is useful in most vulnerable applications where no permanent change is helped about

the host signal . In the literary works, the majority of the planned algorithms tend to be for

digital photos to introduce undetectable information (e.g. [1]–[8]) or maybe a seen

watermark.(e.g [9]).

Inorder to evaluate the performance of the RDH algorithm we usually use hiding

rate and marked image quality as the two important metrics. It is often identified that the

distortion of the image is proportionate to the hiding rate. Inorder to measure this distortion

the Peak Signal to Noise Ratio (PSNR) value of the marked image is used. In general, if the

image histogram [2] is modified directly, it affects the embedding capacity adversely. Even

if the PSNR value is improved by using prediction error based algorithms, due to the

distortion introduced in the image during the embedding operations hardly improve the

visual quality. With the images with inadequate light, increasing the particular visible

excellent is a lot more essential as compared to preserving the particular PSNR value high.

The security of the data being transmitted across a network has become important

with the increase of internet and other communication methods. The most common method

used is hiding data in a cover media and sending it. The cover media used here in RDH is an

image. It can be text, image, audio or video. After the data is embedded in the cover image,

the embedded message is compressed using the Huffman‟s coding which is further

encrypted using the chaotic encryption technique which uses three keys. To send, store and

receive the data efficiently a compression algorithm helps to a large extend. Once

compression is done, encryption is the process which makes the data embedded and

compressed image secured for being transmitted over the network. To our best knowledge

this is the first system that implements the compressed and highly secured RDH algorithm

with contrast enhancement.

In this system, contrast enhancement and data embedding are done simultaneously,

where the contrast enhancement is obtained by equalising the histogram[10] which modifies

the histogram of pixel values. At first, the highest two peaks of the histogram is selected and

keeping the bins in between the peaks unchanged the outer bins are shifted towards

outward. And the bins are splitted into two adjacent bins. The highest bins are further

chosen in the modified histogram and they are split. This process is repeated until the

required contrast enhancement is obtained. There may occur some overflow or underflow

conditions which is eliminated by pre-processing the image and generating a location map.

This location map is attached with the image so that it can be used to extract the data

effectively at the receiver side. Thus pre-processed and contrast enhanced image is then

compressed using a Huffman coding technique which employs the use of a binary tree. The

compressed image is then selected for encryption using a chaotic encryption technique

2751




which uses three encryption keys for the operation. The system is implemented in two sets

of images which prove that it not only increases the contrast, ie, image quality as well as

improves the sending quality by compression and security by encryption.

The rest of the paper is arranged as follows: Section II describes the proposed

algorithm for RDH with contrast enhancement, Compression and Encryption. Section III

shows the experimental results and Section IV gives the conclusion drawn from the system.

2. RDH Algorithm with Contrast Enhancement followed By Compression and

Encryption A. Embedding data with Histogram Modification

The proposed system is focused primarily on gray scale images. The image histogram is

calculated by counting the number of pixels with a gray level value, j, in a given 8-bit gray-level

image where j {0,1,2,….254, 255}. hI (j) denotes the number of pixels in the image with a pixel

value j, where hI denotes the image histogram. Assume I consists of N various pixel values. The

highest two bins are chosen from the N non empty bins in hI. The smaller bin is represented as IS and

the higher bin is represented as IR. The data embedding for a particular pixel with value i in hI is:

i -1 , for i < IS

IS - bk , for i = IS

i‟‟ = i , for IS < i < IR (1)

IR + bk, for i = IR

i + 1, for i > IR

i‟ represents the modified pixel value and bk represents the k-th bit of the message which can be

either 0 or 1. The Eq. (1) is applied to each pixel in hI such that the bins between the peaks remain

unchanged and the outer peaks are moved outward. Thus they can be split two bins ie, IS -1, and IS ,

IR , IR+1.

To retain the the original IS and IR values, 16 pixels in I are excluded from histogram

computing and the LSB of those pixels are included in the binary values to be hidden. While

retrieving data from the image, the peak values are retrieved and the histogram of the marked image

excluding the 16 pixels is calculated. Then to obtain the data bits, Eq (2) is applied to it :

1, if i‟ = IS -1

b‟k = 0, if i‟ = IS (2)

0, if i‟ = IR

1, if i‟ = IR+1

b‟k represents the k-th data bit extracted from the marked image I‟.The reverse of Eq. (1) is applied

on the image to obtain its original values:

2752




i‟+ 1, for i‟ < IS -1

i= IS , for i‟ = IS -1 or i‟ = IS

IR , for i‟ = IR or i‟ = IR+1 (3)

i‟ – 1, for i‟ > IR+1

The excluded 16 pixels are obtained, and restored to recover the original image, from the extracted

binary values.

B. Pre – Process to avoid Overflow and Underflow

All the pixel values are necessarily to be within the range {1……254}. If any of the pixel

value is out of this range (0 or 255) then overflow or underflow condition will occur . To avoid such

error conditions, the image is preprocessed before its histogram is being modified. In pre processing

step, the pixels with value 0 is added by 1 and pixels with value 255 are subtracted by 1 which will

avoid the overflow and underflow condition since while shifting histogram, the possible change for a

pixel value is 1 . While extracting the data, the original value of these pixels is needed and hence

to memorize those values, a location map is generated which has the same size as that of the original

image. The location map assigns a value 1 to the location of modified pixels whereas a value 0 to the

location of unchanged pixels. The location map is pre-computed and added to binary values which

are to be hidden in the image. While recovery operations, the location map is extracted and the

changed pixels are given its original values and then the data is extracted as well as the image is

recovered completely.

C. Enhancing the Contrast

To increase the rate of hiding of the data in the image, the Eq. (1) is repeated several times

on the image such that the peaks are obtained each time an operation is performed and the Eq (1) is

applied to the peaks obtained from the modified histogram. This process helps in achieving

Histogram equalization which increases the contrast of the image. Hence by applying such an

operation Histogram equalization and contrast enhancement are obtained at the same time. Let the

level of equalization be L , ie, the equation is applied L times, then overflow underflow conditions

may occur. To avoid this, while preprocessing, all the pixel values from 0 to L – 1 are added by L

and all the pixel values from 256 – L to 255 are subtracted with L and the corresponding location

map is generated by adding 0s and 1s at the appropriate positions. While extraction the location map

is extracted and the pixel values are replaced with its original values.

D. Huffman’s Encoding

Compression and decompression are the processes done to make efficient, the sending

,storing and receiving of the data. The preprocessed , data embedded image is compressed before it

is subjected to encryption operation. The compression here is obtained by means of Huffman‟s

encoding algorithm which uses binary trees for the purpose.

In an image, each of the pixel is represented using 8 bits. Huffman encoding technique

compressed the image by reducing the number of pixels used to represent a bit. The technique

applies lesser number of bits to the most often appearing pixels and more number of bits to the less

often appearing pixels. Thus, the size of the image thus obtained will be lesser than that of the

original image.

2753




Let us consider the 6x6 image shown below where every pixel is encoded using 8 bits.

Storing this image would require 36x8= 288 bits. Since pixel values are between 0 and 7, we could

assume 3 bits per pixel only; in this case, storing the image would require 36x3=108 bits. Using

Huffman coding, however, it would require 93 bits.

To compress an image, we need a table of bit encodings, (e.g., an ASCII table, or a table

giving a sequence of bits that's used to encode each pixel value). This table is constructed from a

coding tree using root-to-leaf paths to generate the bit sequence that encodes each pixel value.

Figure 2 shows an example of a Huffman coding tree and the corresponding table of bit encodings.

f(0) = 1

f(1) = 4

f(2) = 2

f(3) = 3

f(4) = 2

f(5) = 12

f(6) = 10

f(7) = 2

Figure 1: A simple 6x6 image, its pixel frequencies, and its histogram.

E. Generation of Table of Bit Encodings

In a coding tree, pixel values are stored at the leaves of the tree. Also, a left-edge is labeled 0

(or 1) and a right-edge is labeled 1 (or 0). The Huffman code for any pixel value can be obtained by

following the root-to-leaf path (i.e., leaf corresponding to the pixel value) and concatenating the 0's

and 1's in the path. Figure 2 shows an example of a table of bit encodings (right) generated by a

coding tree (left).

F. Encoding/Decoding:

To encode an image, we replace each pixel value by its corresponding bit encoding. Using

the encoding shown in Figure 2, for example, the first row of the image shown in Figure 1 is

encoded as follows: 10011010000001110. To decode a stream of pixel encodings, start at the root of

the encoding tree, and follow a left-branch for a 0, a right branch for a 1. When you reach a leaf,

write the pixel value stored at the leaf, and start again at the top of the tree. This process is repeated

until all bit encodings have been decoded.

4. Generation of Huffman Coding Tree

We now describe the algorithm for constructing an optimal coding tree. Let's assume that

each pixel value has an associated weight equal to the number of times the pixel value occurs in an

image. In Figure 1, for example, pixel value 0 has a weight 1 while pixel value 5 has a weight 12.

When compressing an image we'll need to calculate these weights. Huffman's algorithm assumes

that we're building a single tree from a group (or forest) of trees. Initially, all the trees have a single

node with a pixel value and the character's weight. Trees are combined by picking two trees, and

1 3 1 5 5 7

5 5 5 1 2 1

6 6 4 0 6 2

6 4 6 6 6 6

3 3 5 5 5 5

6 6 5 5 7 5

0 1 2 3 4 5 6 7

5

10

15

2754




making a new tree from the two trees. This decreases the number of trees by one at each step since

two trees are combined into one tree. The algorithm is as follows:

1. Begin with a forest of trees. All trees are one node, with the weight of the tree equal to the weight

of the pixel value in the node. Pixel values that occur most frequently have the highest weights.

Pixel values that occur least frequently have the smallest weights.

2. Repeat this step until there is only one tree: Choose two trees with the smallest weights, call these

trees T1 and T2. Create a new tree whose root has a weight equal to the sum of the weights T1 + T2

and whose left subtree is T1 and whose right subtree is T2.

3. The single tree left after the previous step is an optimal encoding tree.

(a) (b)

Figure 2. (a) Huffman coding tree corresponding to the image shown in Figure 1; (b) Coding of pixel values. Note

that the code length of each pixel is inversely proportional to its frequency.

The tree shown in Figure 2 is an optimal tree for encoding the image shown in Figure 1; that

is, there are no other trees with the same pixel values that use fewer bits to encode the image in

Figure 1 There are other trees that use 92 bits; for example you can simply swap any sibling nodes

and get a different encoding that uses the same number of bits.

5. Chaotic Encryption

To carry out encryption first the data is embedded in the image, then the image is

compressed and then the chaotic encryption technique is used for encrypting the image. The

encryption employs three encryption keys - an 80 bit session key, an initial parameter key and a

control parameter key. Following are the steps involved in the process of encryption:

Step 1: The input image is converted into its corresponding binary format. Let the number of rows

be R and number of columns be C. Then the total number of pixels can be calculated as R×C = N

0 1111

1 100

2 1010

3 110

4 1011

5 00

6 01

7 1110

2755




pixels. The row matrix of the image is obtained and each of the pixel is converted to its

corresponding binary format. It can be represented as:

Pnk =

821

.

.

28

.

.

22

.

.

21

181211

........

.......

.......

NNN ppp

ppp

ppp

(4)

Step 2 : In this step the initial parameter key is computed for which the user has to enter the 80 bit

session key and a key with valuve between 0 and 1. The 80 bit session key is entered as 20

hexadecimal characters.

K = k1k2k3…. k20 (5)

After converting these characters into its binary format a block k of 24 bits k5k6k7k8k9k10, is extracted

from the session key K. The value of k may vary depending on the chaos needed. X01 is calculated as

:

X01 = (k51 × 20+ ... + k54 × 2

3+ k61 × 2

4+ ... + k64 × 2

7+ ... + k101 × 2

20+ ... + k104 × 2

23) / 2

24 (6)

The initial parameter key X(1) is computed using X01 and X02 where X02 is entered by the user.

X(1) = (X01 + X02) mod 1 (7)

Step 3 : After generating the keys the chaotic sequence X1 X2 X3…XN is generated as in (8).

Xi = μ Xi-1 (1 - Xi-1) (8)

The chaotic sequence starts with the initial parameter key X(1). Μ is the control parameter key

entered by the user. The sequence is normalized to the image scale 0 and 255 using :

Xi= [xi-min(xi)/max(xi)] ×255 (9)

Now, each of the pixel value is converted back to its 8 bit binary number which gives the matrix B

B =

821

.

.

28

.

.

22

.

.

21

181211

....

....

....

NNN bbb

bbbbbb

(10)

Step 4 : The cipher text p‟nk is computed for each of the pnk using the Eq. (11)

p‟nk = pnk bnk

(11)

2756




The step is repeated for each row in B and the encrypted matrix P‟nk is obtained :

P‟nk =

821

.

.

28

.

.

22

.

.

21

181211

'......''

'......''

'......''

NNN ppp

ppp

ppp

(12)

The matrix obtained in Eq. (12) is converted to the values in range 0 to 255 and thereby converted to

a two dimensional array with size R×C. The output is the encrypted image which contains the

original message.

6. Decryption Process

The Triple key decryption process requires the three keys which are provided at the time of

encryption i.e., 80 – bit session key, initial parameter key, and control parameter key. Once the

decryption is done using the reverse procedure applied in encryption, the image is decompressed and

the original image and the message bit are separated to its original format.

Figure 3: Chaotic Encryption and Decryption.

Cover Image Data bits

Session key

Initial key

Control Key

Encrypted Message

Session Key

Initial Key

Control Key

Huffman’s coding Decompression

Data embedded

image

Compressed image

Triple key

encryption

Decompressed

Image

Decrypted image

Triple key

decryption

Cover Image Data bits

2757




7. Procedure of the Proposed Algorithm

Assume the level of data embedding is L. Then the procedure for data embedding and encryption

can be described as follows:

1) Pre-processing is done at first as mentioned in section II- C after excluding the 16 pixels and the

location map is generated to store the location of the pixels which are changed.

2) After pre-processing the image histogram after excluding the 16 pixels at the bottom row is

generated.

3) Now data embedding is done by selecting the highest two peaks in the histogram and applying

Eq. (1) to it and repeating the process with each of the highest peaks in the modified histogram until

the level of embedding is satisfied i.e., L times. The embedding level L, location map length, and the

LSB of 16 excluded pixels and the previous peak values are added wit the last peaks to be split.

4) The LSB of the 16 excluded pixels are replaced by the last split peak values which will give the

final marked image.

5) This marked image is then compressed using the Huffman encoding technique and after

compression, the chaotic encryption technique using the 3 keys for encryption is done on the

compressed image.

After the encryption is done the image is send over the network. At the receiver, decryption,

decompression and data extraction is done and the steps are as follows:

1) The encrypted image is decrypted by the same keys applied at the time of encryption and the

decrypted image is decompressed.

2) After decompression, to get the value of last split peaks, the LSB of 16 excluded pixels are

extracted.

3) By applying Eq. (2) the value of L, length of location map, LSB values of excluded 16 pixels, and

previously splitted peak values are obtained by extracting the data embedded in the last split two

peak values. All of the split peak values are obtained by repeating this process.

4) From the location map, the modified pixel locations are identified and they are given the original

values so that the original image can be extracted and recovered. The maximum value of L is given

as 64 to avoid the ambiguity that may arise such that ,when the pixel value is less than 128 it is

subtracted by L and increased by L otherwise.

The whole procedure is shown in Figure 4.

2758




The figure shows all the steps described in the above steps. The image is selected, Prep-

Processed and the location map is generated. Then the histogram is calculated and data is embedded.

After embedding data the compression and encryption is done, this is reversed to extract the data and

original image.

8. Results

The proposed system was applied on two sets of images. The output proves that the system

is efficient and better than the three MATLAB functions available. Figure 5 shows the input image

and Figure 6 gives the image with the secret data.. It is evident from the images that the contrast of

the input image after embedding data is enhanced by a greater value. While embedding the data , the

embedding level was specified and according to the level the peaks will be equalised. The peak

values ie, IS and IR values for each of the embedding level was obtained.

Figure 5: Input Image Figure 6: Image with secret data

Figure 7 shows a comparison of the histograms of the input image and the image with secret

data. It is seen that the histogram after emedding the data is equalized, which is the major reason for

the contrast enhancement of the image provoided. After the image is embedded with the data, it was

encoded using Huffman‟s encoding technique. After the encoding was done at embedding level 2

the encoding length was found to be 7.4792. While at embedding level 3, the length of encoding was

found to be 7.4843 and so on for the image „Lena‟.

Figure 7 : Histogram of the original image and output image

2759




Then the data was retrieved from the image after decoding and decrypting the image using

reverse of data compression and encryption techniques. The obtained original iamge and the input

image are then compared and cross checked that they both are the same.

Figure 8: The retrieved original image

9. Conclusion In this system, a new Reversible Data Hiding (RDH) technique has been proposed along with

compression and encryption techniques to ensure security and efficiency while sending data along a

network. Experimental results show that image contrast can be enhanced simultaneously while

embedding data in the image by performing histogram equalization. On comparing with the existing

three MATLAB functions - adapthisteq, imadjust, and histeq - it is proven that the image visual

quality is better while using the proposed system. The original image can be extracted as it is

without any distortion after extracting the embedded data. Applying this system for medical and

satellite images as well as for colour images will be the future work.

References

[1] J. Tian, “Reversible data embedding using a difference expansion,” IEEE Trans. Circuits Syst. Video

Technol., vol. 13, no. 8, pp. 890–896, Aug. 2003.

[2] Z. Ni, Y. Q. Shi, N. Ansari, andW. Su, “Reversible data hiding,” IEEE Trans. Circuits Syst. Video

Technol., vol. 16, no. 3, pp. 354–362, Mar. 2006.

[3] D.M. Thodi and J. J. Rodriguez, “Expansion embedding techniques for reversible watermarking,” IEEE

Trans. Image Process., vol. 16, no. 3, pp. 721–730, Mar. 2007.

[4] D. Coltuc and J.-M. Chassery, “Very fast watermarking by reversible contrast mapping,” IEEE Signal

Process. Lett., vol. 14, no. 4, pp. 255–258, Apr. 2007.

[5] V. Sachnev, H. J. Kim, J. Nam, S. Suresh, and Y. Q. Shi, “Reversible watermarking algorithm using

sorting and prediction,” IEEE Trans. Circuits Syst. Video Technol., vol. 19, no. 7, pp. 989–999, Jul. 2009.

[6] X. Li, B. Yang, and T. Zeng, “Efficient reversible watermarking based on adaptive prediction-error

expansion and pixel selection,” IEEE Trans. Image Process., vol. 20, no. 12, pp. 3524–3533, Jan. 2011.

[7] Z. Zhao,H.Luo,Z.-M. Lu, and J.-S. Pan, “Reversible data hiding based on multilevel histogram

modification and sequential recovery,” Int. J. Electron. Commun. (AEÜ), vol. 65, pp. 814–826, 2011.

[8] H. T.Wu and J. Huang, “Reversible image watermarking on prediction error by efficient histogram

modification,” Signal Process., vol. 92, no. 12, pp. 3000–3009, Dec. 2012.

[9] Y. Yang, X. Sun, H. Yang, C.-T. Li, and R. Xiao, “A contrast-sensitive reversible visible image

watermarking technique,” IEEE Trans. Circuits Syst. Video Technol., vol. 19, no. 5, pp. 656–667, May 2009.

[10] J. A. Stark, “Adaptive image contrast enhancement using generalizations of histogram equalization,”

IEEE Trans. Image Process., vol. 9, no. 5, pp. 889–896, May 2000.

[11] P. G. Howard, F. Kossentini, B. Martins, S. Forchhammer, and W. J. Rucklidge, “The emerging JBIG2

standard,” IEEE Trans. Circuits Syst. Video Technol., vol. 8, no. 7, pp. 838–848, Jul. 1998.

[12] The USC-SIPI Image Database [Online]. Available: http://sipi.usc.edu/ database/

[13] Kodak Lossless True Color Image Suite [Online]. Available: http:// www.r0k.us/graphics/kodak/

2760




[14] M.-Z.Gao, Z.-G.Wu, and L.Wang, “Comprehensive evaluation forHE based contrast enhancement

techniques,” Adv. Intell. Syst. Applicat., vol. 2, pp. 331–338, 2013.

Biographies

1st Anju Mariyam Zacharia, was born in Kerala, India on 1991 Feb 26. She received her Bachelor of

Technology from Mahatma Gandhi University, India in 2012. At present she is pursuing her Master

of Technology in Computer Science and Engineering from Mahatma Gandhi University.

2nd Prof. Shyjila P.A., was born in Kerala, India on 1980 Oct 5. She received her Bachelor of

Technology from Mahatma Gandhi University .She has completed her Master of Technology from

Kerala University. At present she is employed as Associate Professor at Mahatma Gandhi University

in Computer Science Department.

3rd Karthika J. S. , was born at Kerala, India on 1991 .She received her Bachelor of Information

Technology from Cochin University, India in 2013. At present she is pursuing her Master Of

Technology in Computer Science and Engineering from, Mahatma Gandhi University.

compressed and highly secured reversible image data hiding ... · highly secured reversible image...

Documents