International Journal of Information Technology and Business Management

International Journal of Information Technology and Business Management

29th April 2013. Vol.12 No.1

© 2012 JITBM & ARF. All rights reserved

ISSN 2304-0777

TEXT EMBEDDING BASED ON CONTOURELET TRANSFORMATION COEFFICIENTS

KhalilIbrahimAlSaifMeaad M. Salih

Ass. Prof. Dr.M. Sc. Student

Dept. of Computer Science

College of Computer & Mathematic Science-Mosul Univ./ IRAQ

Email : ;

Abstract— Text hiding in digital image has recently received quite a bit of attention because it is very important in invisible communication. This paper presents a new data hiding technique for embedding text data in image using (NSCT) . The text data is first converted to ASCII format and then represented in binary form and then, added to contourlet coefficients. A high frequency directional band pass of the contourlet transform is selected for data embedding.

The proposed method show that we could successfully embed data in cover_images and extract it with the average embedding capacity of bits per pixel without any error . High capacity can be achieved using this method according to block size. The accuracy of the stego _image with respect to the original one was checked by evaluating PSNR (Peak Signal to Noise Ratio) , MSE and the correlation factor which studied in details .

Keywords: Steganography ,Contourlet Transform ,Stego_Image ,Cover Image .

1

International Journal of Information Technology and Business Management

29th April 2013. Vol.12 No.1

© 2012 JITBM & ARF. All rights reserved

ISSN 2304-0777

1. INTRODUCTION

Steganographyis the art of hiding information imperceptibly in a cover medium. The word "Steganography" is of Greek origin and means "covered or hidden writing". The main aim in steganography is to hide the very existence of the message in the cover medium [1]. The information-hiding process in astenographic system starts by identifying the redundant bitsin cover medium that can be modified without destroying theintegrity of the medium. The embedding process creates astego medium by replacing these redundant bits with datafrom the secret message to be sent. Steganography findsextensive applications in covert communications, authentication, proof of ownership, customer tracing, featuretagging and data embedding.All stenographic algorithms have to comply with a fewbasic requirements. The most important requirement is that asteganographicalgorithm has to be imperceptible[2] . The aim of the steganography methods is to communicate securely in a completely undetectable manner. The current approaches in steganography domain could be divided into two categories of spatial and transformation domains. In spatial domain, there are many strategies of image steganography based on manipulating the least significant bit using direct replacing of least significant bit levels with the bits of secret image.[3] Because of the limitation of the total number of least significant bit levels in a cover image, these methods have appropriate outputs only when the size of secret image is small enough. To reach this goal, the secret image size is not bigger than 25 percent of the cover image size. Nevertheless, when the hidden message is big proportional to cover image, the quality of stegano image is low in these methods. Another group is related to methods that the embedding process is implemented through frequency domain and wavelet transform. For instance the work presented in [4], with using zero-padding features, the authors presented a steganography method based on image Fourier domain .In [5] a frequency domain method was proposed so that embedding is realized in bit planes of sub-band wavelet coefficients obtained by using the Integer Wavelet Transform . The presented work is among the frequency domain methods that are based on storing the secret image data in block to block form. Many methods have been proposed which are based on the block to block storing in frequency domain [3] . In propoed method we storing data in blocks of coefficients in high frequency sub-bands, but we use NonsampledCounterlet Transform (NSCT) .One of the advantages of counterlet transform is the existence of linearly independent sub-bands. This issue decreases the possibility of detection of the stegano image by steganalysis algorithms [3] .

1. RELATED WORK :

Jiri Fridrich presented in 1999 in his research “A New Steganographic Method for Palette-Based Images “ , he proposed new steganographic technique for embedding messages in palette-based images, such as GIF files. This technique embeds one message bit into one pixel (its pointer to the palette). The pixels for message embedding are chosen randomly using a pseudorandom number generator seeded with a secret key. For each pixel at which one message bit is to be embedded, the palette is searched for closest colors. The closest color with the same parity as the message bit is then used instead of the original color[6].

In 2002, Chin-Chen Chang ,Tung-Shou Chen and Lou-Zo Chung presented in their research “A steganographic method based upon JPEG and quantization table modification” , a novel steganographic method based on joint photographic expert group (JPEG) . The proposed method modifies the quantization table first. Next, the secret message is hidden in the cover-image with its middle-frequency of the quantized DCT coefficients modified.Proposed technique provides acceptable image quality and a large message capacity[7].

In 2009 , Hedieh SAJEDI& Mansour JAMZAD presented in their research “ ContSteg: Contourlet-Based Steganography Method “ a new adaptive steganography method based on contourlet transform is presented that provides large embedding capacity with average 0.05 bits per pixel.they called the proposed method ContSteg. In contourlet decomposition of an image, edges are represented by the coefficients with large magnitudes. In ContSteg, these coefficients are considered for data embedding because human eyes are less sensitive in edgy and non-smooth regions of images. For embedding the secret data, contourlet sub bands are divided into 4×4 blocks. Each bit of secret data is hidden by exchanging the value of two coefficients in a block of contourlet coefficients. [8] .

In 2009, Sudeep P.V. & K.A. NavasSheeba V.S. in their research “A Novel Datahiding Method In Spatial Domain “ proposed technique hides ASCII characters in tle least significant digits (LSD) of the image .This investigation mainly focuses on enhancing the capacity of data hiding in medical images ,user has the choice for selecting the number of LSDs of the pixel magnitude . This method achieving enormous capacity without sacrificing the visual quality of the images. 1 MB data can be embedded in 512 ×512 in gray scale image [9].

DiptiKapoorSarmahNehaBajpai presented in 2010 in their research “Proposed System for Data Hiding Using Cryptography and Steganography “ they are developing a system where they develop a new technique in which Cryptography and Steganography are used as integrated part along with newly developed enhanced security module. In Cryptography they are using AES algorithm to encrypt a message and a part of the message is hidden in DCT of an image; remaining part of the message is used to generate two secret keys which make this system highly secured [10].

in 2011 , Malini Mohan & Anurenjan. P. R in their research “A Novel Data Hiding Method in Image using Contourlet Transform“, they applied an encryption algorithm which provides additional security. A digit of data is embedded by modifying the least significant digit of a contourlet coefficient. A high frequency directional pass band from the contourlet transform is selected for data embedding. High capacity can be achieved using this method. The performance of this proposed method in other transform domains like DCT and wavelet transform alse evaluated and compared with that in contourlet domain [2] .

SushilKumar &S.K.Muttoo presented in 2011 a research “Steganography based on Contourlet Transform“The proposed technique uses a self-synchronizing variable length code to encode the original message which has been proved better than Huffman code in terms of power energy. The secret data then is embedded in the high frequency sub-bands obtained by applying CTT( Contourelet Transform) to the cover-image using variable LSB method and thresholding method. The Contourlet transform is more suitable for data hiding applications as Contourlet gives more edges. Experimental results show that the original message and original image both can be recovered form stego-image accurately and show better imperceptibility than the DWT(Disc. WaveletTransform)based method[11].

3. AIM OF PAPER:In this research a new technique will be achieved for data hiding depend on dividing one of the contourlet coefficients to blocks and embedded the message bits in its center.

1. CONTOURLET TRNSFORM :

The Contourlet transform was proposed by M. N. Do and M. Vetterli. The contourlet transform provides a multi scale and multi-directional representation of an image. It consists of a double filter bank structure for obtaining sparse expansions for typical images having smooth contours. In this double filter bank, the Laplacian pyramid (LP) is first used to capture the point discontinuities, and then followed by a directional filter bank (DFB) to link point discontinuities into linear structures. The required number of directions can be specified by the user. Since contourlets gives more edges, it is more suitable for data hiding applications as more data can be hidden in the high frequency regions without perceptually distorting the original image. The overall results in an image expansion using basic elements like contour segments, and, thus, are named contourlets[2].

4.1 Non-SubsampleContourlet Transform (NSCT)

The non-subsample contourlet transform is a new image decomposition scheme introduced by Arthur L.Cunha, Jianping Zhou and Minh N.Do [12] NSCT is more effective in representing smooth contours in different directions of in an image than contourlet transform and discrete wavelet transform. The NSCT is fully shift invariant, Multi scale and multi direction expansion that has a fast implementation. The NSCT exhibits similar sub band decomposition as that of contourlets, but without down samplers and up samplers in it. Because of its redundancy, the filter design problem of nonsubsampled contourlet is much less constrained than that of contourlet The NSCT is constructed by combining nonsubsampled pyramids and nonsubsampled directional filter bank as shown in Figure 1.The non-subsamples pyramid structure results the multi scale property and nonsubsampled directional filter bank results the directional property.

Figure( 1) The nonsubsampled contourlet Transform (a) nonsubsampled filter bank structure that imp-lements the NSCT. (b) Idealized frequency part-itioning obtained withNSCT

Figure 2.the contourlet transform in level one and four directions : a) orginal image b) decomposed image

4.1.1 NonSubsamplePyramids :

The nonsubsamples pyramid is a two channel non subsamples filter bank as shown in Figurers 3(a).The H0(z) is the low pass filter and one then sets H1(z) =1-H0(z) and corresponding synthesis filters G0(z) =G1(z)=1.The perfect reconstruction condition is given by Bezout identity[12],[13],[14]

H0 (z) G0 (z) +H1 (Z) G1 (Z) =1…………… 1

Figure (3) Non subsample pyramidal filters (a). Ideal frequency response of non subsample pyramidal filter(b).The cascading analysis of three stages nonsubsample pyramid by iteration of two channels nonsubsample filter banks.

Multi scale decomposition is achieved from nonsubsample pyramids by iterating the nonsubsample filter banks. The next level decomposition is achieved by up sampling all filters by 2 in both dimensions. The complexity of filtering is constant whether the filtering is with H(z) or an up sampled filter H(z m ) computed using “ a trous ‟ algorithm The cascading of three stage analysis part is shown in Figurer (3)b.

4.1.2Nonsubsample Directional Filter Banks

The directional filter bank (DFB) is constructed from the combination of critically-sampled two-channel fan filter banks and re-sampling operations. The outcome of this DFB is a tree structured filter bank splitting the 2-D frequency plane into wedges. The nonsubsample directional filter bank which is shift invariant is constructed by eliminating the down and up samplers in the DFB.The ideal frequency response of nonsubsample filter banks is shown in Figure (4) a .

Figure ( 4) Nonsubsample directional filter bank (a) idealized frequency response of nonsubsample directional filter bank.(b) The analysis part of an iterated nonsubsample directional bank .

To obtain multi directional decomposition, the nonsubsample DFBs are iterated. To obtain the next level decomposition, all filters are up sampled by a quincunx matrix given by [12],[13]

The analysis part of an iterated nonsubsample filter bank is shown in Figure 4(b).

5. PROPOSED ALGORITHM :

Figure (5) show the block diagram of proposed algorithm. The algorithm contain two phase the first one handle the embedding data inside the image while the second phase cover the extract the embedded data as follow :

Figure (5) block diagram of proposedsystem .

5.1 Embedding Phase:

The embedding process is done in the following steps at the sender side :

Step 1: The image selected is first decomposed using (NSCT) . The result contains a low pass coefficients and many high pass subbands with same size .studied for the best coefficient suitable for embedding , high pass subbands are chosen for embedding the data .

Step 2: The embedded message will be converted to ASCII code , so that it will be represented in binary format .

Step 3: embedding process steps :

a)The identified high pass subband divided into blocks , and prepared that each block will hold one bit from the secret message .

b)each bit ( 0 or 1) of the message to be embedded by adding it’s value to the value of block center .

m= d+ V_center ……………..3

where d is the bit to be embedded and V_center is the value of center of block .The process is continue until the whole data is embedded .

Step4 : reconstruct the image from the processed coefficients to obtain stego_image ( cover image with secret data) .

Step 5: embed the coefficient level ,block size and threshold value and coefficient level in stego_image using LSB method in its center .

5.2 ExtractingPhase :

In Proposed system , steganography method is blind that mean the recipient must have the cover image to be able to extract embedded data . The embedding process is done in the following steps at the recipient side :

Step 1: extract the coefficient level ,block size and threshold value from stego_image which were embedded using LSB method .

Step 2: The recipient received image (stego_image) and then , decomposed it using ( NSCT ) .

Step 3: The cover image decomposed using (NSCT). identified equivalent high frequency subband S_cover ;

Step4: evaluate the different between S_cover and S_stego to get the difference matrix ( D_matrix) .

Step5 : divided ( D_matrix) in step4 into blocks and then treat center of each block by comparing it with threshold value (V_center) . The process is continued until the whole embedded data will extracted .

Step6: the obtained binary sequence will be converted to ASCII code .

5.3 Applied Example :

Figure (6) show the secret message (2340 char ) that hiding in cover Image ( Barbara 512×512 ) shown in figure (7) after converted it to binary sequence .to obtain the stego_image in figure (8) .

we using contourlet a new data hiding technique for embedding text data in image...... ……………………………………………………………………………….. text data in image using contourle.

Figure (6) : the secret message

Fig (7) : Cover image Fig (8) : Stego-image

6. RESULT AND DISCUSSION:

To evaluate the performance of the proposed embedding algorithm experiment was conducted using gray scale images with different size. Simulations were done using MATLAB.

The image is composed to level two using NSCT. The result contains a low_ pass band and many high_ pass subbands. One of the subbands is chosen for data embedding related to it’s features. Proposed algorithm tested using different block sizes to embed data with different threshold values [0.2 ≤ Threshold ≤ 0.9] for extract data, figure(9) show that embedding process with block size (4×4) give high capacity for embedding data without any error more than other sizes ( 5×5 -11×11). Extraction process depends on threshold value , figure (10) show threshold value and Error rate percent in extracted data according to that threshold.It is clear that Error free appear on threshold value (0.5 and 0.55) for all block sizes .

Figure(9) show embedding capacity according to block size

Figure(10) show error rate in extracted data from stego image 512×512

Image features like PSNR, MSE , Correlation are also analyzed between original image and retrieval one . PSNR stands for peak signal to noise ratio, which is a measure for image quality perpose. The PSNR is most commonly used as a measure of quality of reconstruction in image compression etc. It gives the ratio between the maximum possible power of a signal and the power of corrupting noise that affects the fidelity of its representation. It is most easily defined using the mean squared error (MSE) which computed for two monochrome images I and K where one of them is considered as a noisy approximation of the other. Eq. (4,5) shows the MSE and PSNR evaluation .

MSE=…4

Correlation coefficient between cover image and stego image will be calculated by eq. (6) to see how much the retrieval image close to the original one.

Where , = mean of the value in (A) and (B) computes the mean of the values in A and B respectively .

Figure (11) shows the variation of PSNR and MSE and Correlation for different volume of embedded data .

Figure (11) MSE & PSNR & Correlation

7.CONCLUSION:

From the applied example on the proposed idea using the contourlet coefficient, shows that hiding data in the coefficients of the contourlet gives a robust technique for high security .

Segmentation of the cover image in same block size achieved to reachhigh performance of data embedded, which goes to error free between the original message and the retrieved one when the cover image was segmented on 4×4 block size .