An Optimum Method Research Based on Bit Plane Combination of D igital Image and Fingerprint Watermarking＊

ZHAO Huim in, GUO Yizhen,D ING X iaoyan

(1.Guangdong Polytechnic Nor malUniversity,School of Electronic and Information,Guangzhou 510665,China;2.Xi’an University of Posts&Telecommunication,School of Communication and Information Engineering,Xi’an 710121,China)

1 Introduction

In for mation hiding technology is an emerging research area which encompasses application such as copyright protection for digital media,watermarking,fingerprinting,steganography,and data embedding[1].In the past literature on water marking,it is observed that bit plane method is one of the recommended methods of water marking in spatial domain.This method is characterized by spread spectrum and is blind whilewatermark retrieval[2].Optimal implementation of this method maximizes the fidelity and robustness against different attacks.Thismethod is based on the fact that the least significant bit plane of the image does not contain visually significant information.Therefore it can be easily replaced with water mark bits without affecting the quality of original image[3-4].However the survival of the water mark is an open issue and two main drawbacks of inserting watermark in least significant and most significant bits are:

1)If water mark is inserted in least significant bit planes,then the watermark may not survive against coding,channel noise,mild filtering or random bitflipping.

2)On the other hand,if the water mark is embedded in most significant bit plane,watermark survives but image quality is degraded.

Therefore,to get optimal results,in terms of fidelity,robustness,and high embedding capacity,a new bit plane modification method is proposed in this paper.

To overcome above problems,we propose the novel method for image water marking.Proposed method differs in two different ways than the earlier technique of bit plane watermarking[5-6].Firstly,to prove the ownership or identify the owner,most effective digital signature watermark(as fingerprint data)is embedded instead of pseudorandom watermark.Secondly,instead ofLSB,a previous bit to LSB is identified for watermark embedding to avoid the degradation of image and to survive the water mark after different general attacks like various Conditional compression.

2 Proposed Watermarking Algorithm

Without significant loss of generality,we shall focus on water marking still images with 256 gray levels of size 512×512 pixels.To trade off be tween the invisibility and robustness of the water mark,the high-energy sub-band(LL4)is not used.Further more,the coefficients in high frequency subbands(LH1,HL1 and HH1)are not used since they often contain few energy.In other subbands,we group the coefficients corresponding to the same spatial location together.Figure 1 shows an example of a group with one coefficient from HL4,4 coefficients from HL3,and 16 coefficients from HL2[7-9].

Fig.1 A group ofmultiwavelet coefficients and Bit plane representation

LetX(m,n)be the grey level image and W(m,n)be the fingerprint image, The grey level image is trans for med into the water marked imageYw(m,n)by DMWTwith fingerprint dada embedded.The grey scale imageXis defined as follows:X={X(m,n),m∈{1,...,M},n∈{1,...N}},and M,Naremaximum dimensions of an image,whereX(m,n)∈{0,.......,..255}total number of greylevels[10-11].Step by step algorithm for proposed method is explained below:

Step 1:Decompose the grey level image to bit planes:Grey level image is decomposed in to bit plane image.Each pixel in the image is represented by 8-bits.Therefore the image is decom-posed into eight 1-bit planes,ranging from 8thbit plane for LSB to 1stbit plane for the MSB.The 8thbit plane contains all the lowest order bits in the pixels comprising the image and

1stbit plane contains all the higher order bits as shown in Fig.1 .

The imagewould be decomposed in to bit planes as follows.A subband coefficient Decomposition of original image into 8-bit planes:

Similarly,decomposition of fingerprint image into 8-bit planes:

Whereland kindicates number of bit planes of image and∈{b1,b2,.......b8}.

Step 2:Replace the significant bitplane of original image with fingerprint bit plane(this is done so as to decide by HVS),which bit planes of the image are good for water marking[12].Following set of equations display replacement of 7thbit plane of original image with 1stbit plane of fingerprint data as an example.The same procedure can be adopted for the remaining bit planes of the image.

Combination water marked image is as follows:This bit plane wate rmarked imageYw(m,n)is recomposed in to grey level imageI(m,n)by IDMWT.

Step 3:Formulation for watermarked image subjected to attacks:In real life when watermarked image is distributed on the World Wide Web,it is encountered by different attacks.In this step,watermarked image is subjected to different types of attacks,leading to attacked image as(m,n),i∈difference attacks.

Step 4:Watermark Retrieval:In this step attacked image(m,n)is again transformed in to binary image i.e.8-bit planes as shown below.

Extract the watermark bit plane from the attacked image.This retrieved wate rmark after attack,is denoted as(m,n).

Step5:Computation of CRC:Correlation coefficient be tween retrieved watermark and original fingerprint is estimated using a standard equation(6)[4].The estimated correlation coefficients are denoted as CRCi(l,k).Where,iindicate different attacks,lis taken as 7thand 8thbit planes of original image as selected in step 2 and kdenotes the bit planes of fingerprint watermark from 1 to 8.The quality of watermarked image is observed by HVS.CRC varies between 0 and 1.CRC is defined as given below:

Step6:Estimation of peak signal to noise ratio(PSNR):PSNRis calculated by using following equation.Capacity of the original image to carry the finger print water mark is computed by measuring PSNR,which is defined as follows:

Mean square error is defined as:

Step 7:Weighted correlation coefficient computation:Weighted correlation coefficient is defined as follows:

Where,βiare the different weightings of attacks such that total,andiis the number of attacks in various compression.The identified attacks are assigned weightings based on damage caused,frequency,intensity and criticality or any other such criterion by the user.Based on these weightings,considering all the ten attacks,weighted correlation coefficient are estimated,for each bit plane combination of image and watermark under consideration.The step is repeated for combinations of selected bitplanesof image and the entire bit planes of watermark respectively.

Step 8:Optimization:The above step 7 is repeated by varying the weightings of attacks.The bit plane combination of original image and fingerprint for which,the weighted correlation coefficient is maximum,is selected as the optimized one for the given user requirements.This combination is used for optimized watermarking in terms of robustness and fidelity.

3 Experiment Results and Analysis

Here,we employed Lenna,Baboo,Perppers as test carrier images,which size of the images was fully 512×512,Size of the embedded fingerprint water marks was 64×64,as figure 2.in the bit planes combination,we use 1stbit plane of fingerprint water mark embedded in 7thbit plane of original image.

The invisibility of the hided image may be measured by histogram statistics charts,and Peak-value SignalNoise Ratio(PSNR)[13-14].To the definite information hiding Capacity,the method hold a kind of different hiding efficiency in various compress ratio with JPEG2000[15].

Retrieval of fingerprint image after difference compression attack was showed in finger 3 when the compression factor Q=100,Q=20,Q=10 andQ=7,and Q=5,as well asQ=2,respectively.The histogram statistics charts of the Lennaimages composited was showed in figure 4,respectively.

Fig.2 The carrier image and embedded fingerprint images

Fig.3 Retrieval of fingerprint image after difference compression attack

Fig.4 The histogram statistics charts of the Lenna images composited data

From these simulation results,we can find out,where the images composited data possessed better image quality,and the influences to the carrier images was a littler after embedded a fingerprint[8].Another,the shapes of the histogram statistics charts closed to auxiliary distributions.It shows,the method has better invisibility in various compression ratio,and the highest compression ratio is able to reach the 100∶5 at conditioned embedding images.

4 Conclusions

Based on bit plane decomposition of the original image and the fingerprint image,the paper proposed an effectively information hiding method,which may adaptively embed hidden fingerprint data into wavelet coefficients from low to high frequency sub-band according to in space bit Planes.Experimentation results shows,that The proposed method is found to be useful for authentication and to prove legal ownership,and possesses better tradeoff be tween invisibility and robustness in various conditional compression.

[1] PRAYOTH K,KITTIA,ARTH IT S.A robust image watermarking scheme using multiwavelet tree[C]∥Proceedings of the World Congress on Engineering,London,U.K.July 2-4,2007.

[2] SUSHMA K,MANESH K.Optimization of bitplane combination for efficient digital image after marking[J].International Journal of Computer Science and Information Security,2009,4(2):19-20.

[3] ZhANG L.Combined the information hiding technology with JPEG2000[D].University of TianJin,2003:95-98.

[4] ZHAO H M.Fingerprint system and DSP fast processing technology[J].Research&Progress of Solid State Electronics,2004,24(3):337-342.

[5] TAUBMAN D S,MARCELL IN M W,et al. Image compression foundation,standardand practice of the JPEG2000[M].Electronics Industrial Publisher,2004:415-417.

[6] L I K,ZHANG X P.An image water marking method integratingwith JPEG-2000 still image compression standard[C]∥Electrical and Computer Engineering IEEE CCECE Canadian Conference on,2003:2053-2054.

[7] TRAPPE W,WU M,WANG Z J.Anti-collusion fingerprinting formultimedia[J].IEEE Transactions on Signal Processing,2008,53(4):1069-1087.

[8] WANG Z J,WU M,TRAPPEW.Group-oriented fingerprinting for multimedia forensics[J].EURASIP Journal on Applied Signal Processing,2004,14(8):2142-2162.

[9] WANG Y,L IANG F,X IAO M M.Color image watermarking adaptively in DC coefficients[J].Acta Scientiarum Naturalium Universitatis Sunyatseni,2010,49(4):43-48.

[10] S WANSON M D,KOHAYASH IM,TEWFIK A.Multimedia data embedding and watermarking technologies[J].Proc of the IEEE,1998,86(6):1064-1087.

[11] REN J,NADOOSHAN T.A cryptographic watermark embedding technique[C]∥IEEE Asilomar Conf on Signals,Systems and Computers,2004:382-386.

[12] MAEDER A J,PLAN ITZ B M.Medical image watermarking formultiple modalities[C]∥34th IEEE Proc on Applied Imagery and Pattern Recognition Workshop,2005:158-165.

[13] FEIC,KUNDER D,K WONG R H.Analysis and design of secure water mark-based authentication systems[J].IEEE Trans OnInfor mation Forensics and Security,2006,1(1):43-55.

[14] PRAYOTH K,KITTIA,ARTH IT S.A new approach for optimization in image water marking by using genetic algorithms[J].IEEE Trans on Signal Processing,2005,53(12):4707-4719.

[15] COX J,K IL IAN J,THOMSON F,et al.Secure spread spectrum watermarking for multimedia[J].IEEE Trans on Image Processing,1997,6(12):1673-1686.