Steganography using Reversible Texture Synthesis based on Error Histogram Shift

IJIRST –International Journal for Innovative Research in Science & Technology| Volume 3 | Issue 04 | September 2016 ISSN (online): 2349-6010

All rights reserved by www.ijirst.org 184

Steganography using Reversible Texture

Synthesis based on Error Histogram Shift

Anumol Antony Dr. Arun Kumar M N

PG Student Assistant Professor

Department of Computer Science & Engineering Department of Computer Science & Engineering

Federal Institute of Science & Technology Federal Institute of Science & Technology

Abstract

Steganography is the method for concealing data inside another file, message, image, or video. The purpose of steganography is

to hide data in a manner that existence of communication is unknown by an attacker. This proposed work presents stegnography

in texture images utilizing reversible texture synthesis based on error histogram shift. Texture synthesis process synthesizes a

large texture image from a smaller texture image, which has same local appearance. The texture synthesis procedure is fabricated

into steganography concealing secret messages and in addition the source texture. The algorithm conceals the source texture

image and embeds the secret messages through the procedure of texture synthesis and error histogram shift. This permits us to

extract the secret messages and the source texture from a stego synthetic texture.

Keywords: Steganography, Reversible Texture Synthesis, Texture Synthesis, Error Histogram Shift, Stego Synthetic

Texture

_______________________________________________________________________________________________________

I. INTRODUCTION

Steganography is the method of hiding a message, file, image, or video within another file, message, image, or video. The word

steganography combines from the two Greek words steganos means protected, and grapheins means writing. The advantage of

steganography than cryptography is that the secret message does not attract the attention of the attackers by simple observation.

The cryptography protects only the content of the message, while steganography protects the both messages and communication

environment.

Most stenographic methods take over an existing image as a cover medium. When embedding secret messages into this cover

image, distortion of the image may occur. Because of this reason two drawbacks occur. First, since the size of the cover image is

fixed, the more secret messages which are embedded leads to more image distortion. Therefore to maintain image quality it will

provide limited embedding capacity to any specific cover image. Second, that image steganalysis approach is used to detect

hidden messages in the stego image. This approach can defeat the image steganography and reveals that a hidden message is

being carried in a stego image.

Embedding capacity is one of the most important requirements for steganography methods, and it is important for

steganography process not to leave any noticeable traceable to the human eyes after hiding the secret data. The proposed method

uses error histogram shift. The use of error histogram shift leads to a better embedding capacity.

The method uses a secret key for source texture recovery. Secret Key Steganography is defined as a steganographic system

that requires the exchange of a secret key (stego-key) prior to communication. The source texture is embedded using the secret

key. Only the parties who know the secret key can reverse the process and recover the source texture. Here a perceived invisible

communication channel is present.

This steganography method exchanges a stego-key, which makes it more susceptible to interception. Our approach offers three

advantages. First, since the texture synthesis can synthesize an arbitrary size of texture images, the embedding capacity which

the scheme offers is proportional to the size of the stego texture image. Secondly, a steganalytic algorithm is not likely to defeat

this steganographic approach since the stego texture image is composed of a source texture rather than by modifying the existing

image contents.

Third, the reversible capability inherited from the scheme provides functionality to recover the source texture. Since the

recovered source texture is exactly the same as the original source texture, it can be employed to proceed onto the second round

of secret messages for steganography if needed.

Experimental results have verified that the proposed algorithm can provide more embedding capacities, produce visually

plausible texture images, and recover the source texture. Theoretical analysis indicates that there is an insignificant probability of

breaking down the steganographic approach, and the scheme can resist an RS steganalysis attack.

The rest of this paper is organized as follows. In section II, literature survey is briefly described. Section III describes the

methodology. In section IV presents the experimental results and analysis and finally section V summarizes the system.

Steganography using Reversible Texture Synthesis based on Error Histogram Shift (IJIRST/ Volume 3 / Issue 04/ 033)


II. RELATED WORKS

In [1] A. A. Efros proposed a non-parametric method for texture synthesis. The texture synthesis process grows a new image

outward from an initial seed, consider one pixel at a time. First, chose a single pixel so that the model captures high frequency

information as possible. All previously synthesized pixels in a square window around single pixel are used as the context. Using

the probability tables for the distribution of single pixel, synthesis is proceeded, given all possible contexts. An approximation

can be obtained by using various clustering techniques. Instead, for each new context, the sample image is queried and the

distribution of the single pixel is constructed as a histogram of all possible values that occurred in the sample image.

In [2] L.-Y. Wei and M. Levoy present an efficient algorithm that can efficiently synthesize a wide variety of textures. The

algorithm is easy to use and it generates textures with perceived quality equal to or better than those produced by previous

techniques, but runs two orders of magnitude faster. The inputs consist of an example texture patch and a random noise image

with size specified by the user. The algorithm modifies this random noise to make it look like the given example. The algorithm

is derived from Markov Random Field texture models and generates textures through a deterministic searching process. This

synthesis process is accelerated using tree-structured vector quantization.

In [3] Liang et al. introduced the patch-based sampling strategy. The algorithm synthesizes textures from an input sample.

This patch-based sampling algorithm is very fast and it creates high-quality texture image. This algorithm works well for a wide

variety textures likes regular to stochastic textures. The patches are sampled according to a nonparametric estimation of the local

conditional MRF density function. This avoids mismatching features across patch boundaries. The building blocks of the patch-

based sampling alg rithm are patches of the input sample texture to construct the synthesized texture. Patch-based sampling

algorithm combines the nonparametric sampling and patch pasting strengths .The texture patches in the sampling scheme provide

implicit constraints to avoid garbage found in some textures.

In [4] Efros and Freeman proposed a method that generates a new texture by stitching together small patches of existing

textures. This process is known as image quilting. It is very fast and simple texture synthesis algorithm. The generalization of the

method is used to perform texture transfer. In the quilting algorithm first go through the image to be synthesized in raster scan

order in steps of one block (minus the overlap). For every location, search the input texture for a set of blocks that satisfy the

overlap constraints (above and left) within some error tolerance. Randomly pick one such block. Compute the error surface

between the newly chosen block and the old blocks at the overlap region. Find the minimum cost path along this surface and

make that the boundary of the new block. Paste the block onto the texture. This step is repeated. This method is extended to

perform texture transfer.

In [5] K. Xu et al. explore the use of salient curves in synthesizing intuitive, shape-revealing textures on surfaces The texture

synthesis is guided by two principles: matching the direction of the texture patterns to those of the salient curves, and aligning

the prominent feature lines in the texture to the salient curves exactly. This is called feature-aligned shape texturing. The

technique is fully automatic, and introduces two novel technical components in vector-field-guided texture synthesis: an

algorithm that orients the salient curves on a surface for constrained vector field generation, and a feature-to-feature texture

optimization.

In [6] M. F. Cohen proposed a simple stochastic algorithm for non-periodically tile the plane with a small set of Wang Tiles

for image and texture generation at runtime. Wang Tiles are squares in which each edge is assigned a color. A valid tiling

requires all shared edges between tiles to have matching colors. The main advantage of using Wang Tiles is that once the tiles

are filled, large expanses of non-periodic texture (or patterns orgeometry) can be created as needed very efficiently at runtime. If

the set of tiles is rich enough there Is no periodicity. We can fill inside the tiles anything we want such as texture, geometric

primitives or points to create Poisson distribution. The user fills in the Wang tiles on her own. The system displays the result of

the tiling interactively. The generation of large textures is very fast.

In [7] Ni et al. proposed a novel reversible data hiding algorithm, which can recover the original image without any distortion

from the marked image after the hidden data have been extracted for embedding the data into the image. Histogram shifting is a

preferred technique among existing approaches of reversible image data hiding because it can control the modification to pixels,

thus limiting the embedding distortion, and it only requires a small size location map, thereby reducing the overhead

encountered. To embed data into the image, the algorithm utilizes the zero or the minimum points of the histogram of an image

and slightly modifies the pixel gray scale values. It can embed more data than any other existing reversible data hiding

algorithms. The algorithm applicable to a wide range of images such as commonly used images, medical images, texture images,

aerial images and all of the 1096 images in CorelDraw database.

In [8] C. Han developes a multiscale texture synthesis algorithm. A novel example-based representation, called an exemplar

graph is proposed that simply requires a few low-resolution input exemplars at different scales. Exemplar graph is an input

representation better suited for the multiscale setting. The nodes in the graph are exemplars, and they are connected by directed

and weighted edges. Moreover, by allowing loops in the graph, we can create infinite zooms and infinitely detailed textures that

are impossible with current example-based methods. Example-based texture synthesis algorithms have gained widespread

popularity for their ability to take a single input image and create a perceptually similar non-periodic texture. However, previous

methods rely on single input exemplars that can capture only a limited band of spatial scales. For example, synthesizing a

continent-like appearance at a variety of zoom levels would require an impractically high input resolution.

In [9] H. Otori and S. Kuriyama proposes a new type of image coding method using texture image synthesis. A digital camera

mounted on a mobile phone is utilized as a data input device to obtain embedded data by analyzing the pattern of an image code



such as a 2D bar code. Regularly arranged dotted-pattern is first painted with colors picked out from a texture sample, for having

features corresponding to embedded data. Our texture synthesis technique then camouflages the dotted-pattern using the same

texture sample while preserving the quality comparable to that of existing synthesis techniques. The textured code provides the

conventional bar code with an aesthetic appeal and is used for tagging data onto real texture objects, which can form a basis for

mobile data communications. This improved the quality of data-embedded textures.

In [10] Yimo Guo proposed video texture synthesis with multi-frame LBP-TOP and diffeomorphic growth model. Two key

factors, such as frame representation and blending artifacts, that affects the synthesis performance. To improve the synthesis

performance from two features: First, effective frame representation is used to capture both the longitudinal information in

temporal domain and the image appearance information in spatial domain. Second, artifacts that reduce the synthesis quality are

significantly suppressed on the basis of a diffeomorphic growth model. The proposed video texture synthesis approach has

mainly two stages such as video stitching stage and transition smoothing stage. In the video stitching stage, a video texture

synthesis model is proposed to generate an infinite video flow. This proposed method presents a new spatial temporal descriptor

to give an effective representation for different types of dynamic textures. In the second stage of video synthesis, a smoothing

method is presented to improve synthesis quality. It aims to set up a diffeomorphic growth model to emulate local dynamics

around stitched frames.

In [11] Kuo-Chen Wu and Chung-Ming Wang proposed an approach for steganography using a reversible texture synthesis. A

texture synthesis process synthesizes a new texture image from a smaller texture image which has a similar local appearance and

an arbitrary size. This method combines the texture synthesis process with steganography to conceal secret messages. This

scheme offers many advantages. First, the embedding capacity is proportional to the size of the stego texture image. Second,

steganalytic algorithms not defeat this steganographic approach. Third, this allows recovery of the source texture.

III. METHODOLOGY

The proposed approach steganography using reversible texture synthesis is used for hiding the secret messages. Steganography is

the method of hiding a message, file, image, or video within another file, message, image, or video. A texture synthesis creates

large texture image from small texture image with a similar local appearance and arbitrary size. This method combines the

texture synthesis process and steganography for concealing secret messages as well as the source texture.

The basic unit used for the steganographic texture synthesis is referred to as a patch. A patch represents an image block of a

source texture where its size is user-specified. Fig.1(a) illustrates a diagram of a patch. We can denote the size of a patch by its

width (Pw) and height (Ph). A patch contains the central part and an outer part where the central part is referred to as the kernel

region with size of Kw ×Kh, and the part surrounding the kernel region is referred to as the boundary region with the depth (Pd).

Fig. 1: (a) Patch (b) Kernel blocks (c) Source patch (d) Boundary mirroring and expanding for a source patch.

A source texture with size of Sw × Sh can be subdivided into a number of non-overlapped kernel blocks, each of which has

the size of Kw × Kh, as shown as Fig.1(b). Let KB represent the collection of all kernel blocks thus generated, and ǁKBǁ

represent the number of elements in this set. The indexing for each source patch kbi is employed as KB ꞊ { kbi │i = 0 to ǁKBǁ -

1} . As an example, given a source texture with the size of Sw × Sh = 128 × 128, if we set the size Kw × Kh as 32 × 32, then we

can generate ǁKBǁ = 16 kernel blocks. Each element in KB can be identified as {kb0, kb1, . .. , kb15}.

We can expand a kernel block with the depth Pd at each side to produce a source patch. The expanding process will overlap its

neighbor block. Fig. 1(c) indicates the boundary region of source patch sp4 when we expand the kernel block kb4 to overlap the

kernel blocks kb0, kb1, kb5, kb8, and kb9. If a kernel block is located around the boundary of a source texture, we operate the

boundary mirroring using the kernel blocks symmetric contents to produce the boundary region, as shown in Fig. 1(d) for the

kernel block kb4.



Similar to the kernel block, we can denote SP as the collection of all source patches and SPn = ǁSPǁ as the number of elements

in the set SP. The indexing for each source patch spi is employed as SP = {spi | i = 0 to ǁSPǁ - 1}.

Given a source texture with the size of Sw x Sh, we can derive the number of source patches SPn using (1) if a kernel block

has the size of Kw x Kh. In the proposed method, we assume the size of the source texture is a factor of the size of the kernel

block to ease the complexity.

SPn = 𝑆𝑤

𝐾𝑤 x

𝑆ℎ

𝐾ℎ (1)

Our steganographic texture synthesis algorithm needs to generate candidate patches when synthesizing synthetic texture. The

concept of a candidate patch is trivial: we employ a window Pw x Ph and then travel the source texture (Sw x Sh) by shifting some

pixel each time following the scan-line order. Let CP = {cpi │ i = 0, 1, . . . , CPn-1} represent the set of the candidate patches

where CPn = ǁCPǁ denotes the number of elements in CP.

When generating a candidate patch, we need to ensure that each candidate patch is unique; otherwise, we may extract an

incorrect secret message. In the method, we employ a flag mechanism. We first check whether the original source texture has

any duplicate candidate patches. For a duplicate candidate patch, we set the flag on for the first one. For the rest of the duplicate

candidate patches we set the flag off to ensure the uniqueness of the candidate patch in the candidate list. The method has two

procedures such as

– Message embedding procedure

– Message extracting procedure

Message Embedding Procedure

In message embedding procedure, first divides the source texture image into image block, called patches. To produce an index

table for recording the location of the corresponding source patch. Establish a blank image as workbench where its size is equal

to the synthetic texture. Then paste the source patches into workbench by referring the source patch IDs stored in the index table

to produce a composition image. Then find Mean square error of overlapped region between the synthesized area and the patch

which want to place. Ranking these patches based on increasing order of Mean Square Error. Then select patches from list where

its rank equals the decimal value of an n-bit secret message.

The message embedding procedure involves mainly three steps, shown in fig.2. They are:

– Index Table Generation

– Patch Composition Process

– Message Embedding

Fig. 2: Flowchart of message embedding procedure

Index Table Generation

The first process of this work is the index table generation where an index table is created to preserve the location of the source

patch set SP inside the synthetic texture. The index table will allow us to access the synthetic texture and extract the source

texture completely. The texture of any size according to user's wish can be generated using this index table.



Fig. 3: An illustration of composition image

The dimension of the index table (Tpw × Tph) is first determined. Given the parameters Tw and Th, which are the width and

the height of the synthetic texture we intend to synthesize, the number of entries in this index table can be determined using

equation (2) where TPn denotes the number of patches in the stego synthetic texture.

TPn = TPw × TPh = Tw−Pw

Pw−Pd+ 1˩ ×

Th−Ph

Ph−Pd+ 1˩ (2)

Inorder to achieve the manner of reversibility during the distribution of the source texture, we avoid positioning a source

texture patch on the borders of the synthetic texture. The first-priority position L1 and the second-priority position L2, for two

types of priority locations where ǁ L1 ǁ and ǁ L2 ǁ, derived in (4), represent the number in the first-priority and second-priority

positions, respectively.

ǁL1ǁ = 𝑇𝑝𝑤−2

2˥ ×

𝑇𝑝ℎ−2

2˥ (3)

ǁL2ǁ = 𝑇𝑝𝑤−2

2˥ ×

𝑇𝑝ℎ−2

2˥ (4)

Patch Based Composition

The second step is to attach the source patches into a workbench to create a composition image. First set up an empty image as

the workbench where the size of the workbench is proportional to the synthetic texture. By referring to the source patch IDs

stored in the index table, we then attach the source patches into the workbench. During the attaching process, if no imbrications

of the source patches are found, we can attach the source patches directly into the workbench. However, if pasting locations

cause the source patches to overlap each other, we employ the image quilting technique to reduce the visual artifact on the

overlapped area.

Message Embedding

The secret message is embedded into the stego synthetic texture by the error histogram shift. At the beginning, one of the matrix

of the stego synthetic texture is saved and LSBs is cleared. The message is converted into ASCII values and then again converted

into binary. The LSBs along with the messages is embedded using the error histogram shift. For that the histogram of

interpolation errors is calculated. First divide the histogram of estimating errors into two parts, i.e., the left part and the right part,

and search for the highest point in each part, denoted by LM and RM respectively. Search for the zero point in each part, denoted

by LN and RN. To embed messages into positions with an estimating error that is equal to RM, shift all error values between

RM+1 and RN-1 with one step toward right, and then, we can represent the bit 0 with RM and the bit 1 with RM+1. The

embedding process in the left part is similar except that the shifting direction is left, and the shift is realized by subtracting 1

from the corresponding pixel values.

The overflow/underflow problem occurs when natural boundary pixels change from 255 to 256 or from 0 to -1. To avoid it,

we only embed data into estimating error with its corresponding pixel valued from 1 to 254. However, ambiguities still arise

when nonboundary pixels are changed from 1 to 0 or from 254 to 255 during the embedding process. These created boundary

pixels in the embedding process are defined as pseudo-boundary pixels. Hence, a boundary map is introduced to tell whether

boundary pixels in marked image are natural or pseudo in extracting process. It is a binary sequence with bit “0” for natural

boundary pixel, bit “1” for pseudo-boundary pixel.

The LSB saved, LMs, LNs, RMs, RNs, length of data into marginal pixels are embedded using the LSB embedding. Then the

histogram of interpolation error is calculated and the message is embedded using the error histogram shift.

Message Extracting Procedure

The message extracting for the receiver side involves extracting the secret message concealed in the stego synthetic texture,

generating the index table, retrieving the source texture. The message extracting procedure involves mainly two steps as shown

in fig.4. They are:

Secret Message Extraction

Source Texture Recovery



Fig. 4: Flowchart of message extraction procedure

Secret Message Extraction

The stego synthetic texture is received at the receiver and using the error histogram shift the message is extracted from stegno

synthetic texture. For this, initially the matrix in which the data is embedded is extracted from the stego synthetic texture image.

Next LMs, LNs, RMs, RNs and the length of data are extracted from it. Then the LSBs are recovered that is saved during the

embedding. After that the message is extracted using the error histogram shift.

Source texture recovery

Using the secret key, the same index table is generated as in the embedding procedure. The next step is the source texture

recovery. Each kernel region with the size of Kw × Kh and its corresponding order with respect to the size of Sw × Sh source

texture can be retrieved by referring to the index table with the dimensions Tpw × Tph. We can then arrange kernel blocks based

on their order, thus retrieving the recovered source texture which will be exactly the same as the source texture.

IV. EXPERIMENTAL RESULT

The method was implemented in MATLAB 2012 prototype. The proposed work was performed on a desktop PC with the

following characteristics: Intel Core i3 CPU, 3.4 GHz, 4 GB RAM. The experiment was implemented using some source

textures. To validate the developed method, the computational results obtained by implementing the developed method is

evaluated and compared with the methods presented by earlier researchers.

The proposed method is demonstrated by a texture image. The example demonstrates the message embedding and message

extraction. In message embedding, first the mirroring of source texture is done. It is shown in fig.5.

Fig. 5: Boundary Mirroring



Then an index table is created to preserve the location of the source patch set inside the synthetic texture. The index table will

allow us to access the synthetic texture and extract the source texture completely. The source patches are attached into a

workbench to create a composition image.

Fig. 6: Index table and the composition image generated

Then the image quilting technique is performed to reduce the visual artifact on the overlapped area. Then the secret message

is embedded into the stego synthetic texture by the error histogram shift.

In extracting procedure, the message extracting for the receiver side involves extracting the secret message concealed in the

stego synthetic texture, generating the index table, retrieving the source texture. Initially message is extracted from the stego

synthesis texture using the error histogram shift. Then the source patches are recovered using the index table and the secret key.

We adopt two source textures for the results of our collection. Table-1 presents the total embedding capacity that the algorithm

can provide when different resolutions of the synthetic texture. It is interesting to point out that given a fixed number of BPP, the

larger the resolutions of the source texture Sw × Sh (96 × 96 vs. 128 × 128), smaller the total embedding capacity (TC) the

algorithm will offer (5992 bits vs. 4778 bits). This is because the larger source texture will contain more source patches SPn (9

vs. 36) that we need to paste which cannot conceal any secret bits, thus reducing the total embedding capacity. We can increase

the embedding capacity by embedding data in all the 3 matrices. We compare the embedding capacity with the work presented

by Kuo-Chen Wu and Chung-Ming Wang. The source texture images of 96 × 96 pixels embed 712 bits and the proposed work

embed minimum of 5992 bits. Besides, the scheme extracts the secret messages correctly. Table – 1

Comparison of embedding capacity

To compare the original source texture and the one that is retrieved in the extraction process, PSNR value is calculated. The

PSNR block computes the peak signal-to-noise ratio, in decibels, between two images. This ratio is often used as a quality

measurement between the original and a compressed image. The higher the PSNR, the better the quality of the compressed, or

reconstructed image.



Fig. 7: PSNR value showing infinity

The fig.7 shows that the PSNR value is infinity. It means that the quality of image is very high i.e the source texture has been

recovered successfully.

V. CONCLUSION

This work proposes a reversible steganographic algorithm using texture synthesis based on error histogram shift. Given an

original source texture, first we have to produce a large stego synthetic texture hiding the secret messages. By using a

conventional patch-based method the textures are synthesized. The proposed method also provides reversibility to retrieve the

original source texture from the stego synthetic textures, making possible a second round of texture synthesis if needed. This

steganography method based on error histogram shift minimizes the possible distortion during the embedding process to

minimize the probability of discovering the secret message data from unauthorized users and also resulting in high embedding

capacity.

REFERENCES

[1] A. A. Efros and T. K. Leung, “Texture synthesis by non-parametric sampling", in Proc. 7th IEEE Int. Conf. Comput. Vis., Sep. 1999, pp. 10331038.

[2] L.-Y. Wei and M. Levoy. “Fast texture synthesis using tree-structured vector quantization", in Proc. 27th Annu. Conf. Comput. Graph. Interact. Techn., 2000, pp. 479488.

[3] L. Liang, C. Liu, Y.-Q. Xu, B. Guo, and H.-Y. Shum, “Real-time texture synthesis by patch-based sampling", ACM Trans. Graph., vol. 20, no. 3, pp.

127150, 2001. [4] A. A. Efros and W. T. Freeman, “Image quilting for texture synthesis and transfer", in Proc. 28th Annu. Conf. Comput. Graph. Interact. Techn., 2001, pp.

341346.

[5] K. Xu et al., \Feature-aligned shape texturing", ACM Trans. Graph., vol. 28, no. 5, 2009, Art. ID 108. [6] M. F. Cohen, J. Shade, S. Hiller, and O. Deussen, “Wang tiles for image and texture generation", ACM Trans. Graph., vol. 22, no. 3, pp. 287294, 2003.

[7] Z. Ni, Y.-Q. Shi, N. Ansari, and W. Su, “Reversible data hiding", IEEE Trans. Circuits Syst. Video Technol., vol. 16, no. 3, pp. 354362, Mar. 2006.

[8] C. Han, E. Risser, R. Ramamoorthi, and E. Grinspun, “Multiscale texture synthesis", ACM Trans. Graph., vol. 27, no. 3, 2008, Art. ID 51. [9] H. Otori and S. Kuriyama, “Texture synthesis for mobile data communications"' IEEE Comput. Graph. Appl., vol. 29, no. 6, pp. 7481, Nov./Dec. 2009.

[10] X. Li, B. Li, B. Yang, and T. Zeng, “General framework to histogram shifting- based reversible data hiding", IEEE Trans. Image Process., vol. 22, no. 6,

pp. 21812191, Jun. 2013. [11] Kuo-ChenWu and Chung-MingWang, “Steganography Using Reversible Texture Synthesis IEEE Trans.on Image Processing", IEEE Trans.on Image

Processing, VOL. 24, NO. 1, Jan 2015.

Steganography using Reversible Texture Synthesis based on Error Histogram Shift

Education

Transcript of Steganography using Reversible Texture Synthesis based on Error Histogram Shift