250 IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 Reversible Image Watermarking using Bit Plane Coing an Lifting Wavelet Transform S. Kurshi Jinna, Dr. L. Ganesan PET Engineering College /Professor & Hea of the Department of Computer Science an Engineering, Vallioor, Inia A.C College of Engineering an Technology / Professor & Hea of the Department Computer Science an Engineering, Karaikui, Inia Summary This paper proposes a istortionless image ata hiing algorithm base on integer wavelet transform that can hie ata into the original image.the ata can be retrieve an the original image can be recovere without any istortion after the hien ata are extracte. This algorithm hies ata into one or more mile bit-plane(s) of the integer wavelet transform coefficients in the LH, HL an HH frequency sub bans. It can embe more ata into the bit planes an also has the necessary imperceptibility requirement. The image histogram moification may be use to prevent grayscales from possible overflow or unerflow. Experimental results have emonstrate the performance of the algorithm. Key wors: Reversible image ata hiing, bit plane, compression, integer wavelet transform, lifting scheme. 1. Introuction Many ata embeing methos use proceures that in which the original image is istorte by quite a small amount of noise ue to ata embeing itself. This istortion cannot be remove completely ue to quantization, bit-replacement, or truncation at the grayscale ens. Even though the istortion is often quite small, it may not be acceptable for meical imaging for legal reasons or for military images inspecte uner altere viewing conitions like filtering or zooming. In this paper, we introuce a approach for high-capacity ata embeing that is lossless without any istortion. After the embee information is extracte from the stegoimage, we can revert to the exact copy of the original image before the embeing occurre. The new metho can be use as a powerful tool to achieve a variety of tasks that nees istortion-free image after watermark embeing an extraction of watermarks. The propose concept can be extene to commonly use image formats. Two techniques propose in [1] is base on robust spatial aitive watermarks combine with moulo aition an the secon one on lossless compression an encryption of bit-planes The first technique embes the hash of the whole image as a payloa for a robust watermark an the secon metho for invertible authentication base on lossless compression of bit-planes an encryption is much more transparent for analysis. A high capacity istortionless ata embeing metho is presente which has opene many lossless ata embeing methos [2]. A metho for reversible ata-embeing in igital images using a technique calle ifference expansion is iscusse. Location map is use to locate the marke coefficients. The reunancy in the igital content to achieve reversibility is use. The payloa capacity limit an the visual quality of embee image are consiere [3]. Reversible ata hiing, in which the watermarke image can be reverse to the original cover meia exactly, has attracte increasing interests from the ata hiing community. The existing reversible ata hiing algorithms, have been classifie as those evelope for fragile authentication, for achieving high ata embeing capacity, for semi-fragile authentication. In each category the principles, merits, rawbacks an applications of these algorithms are analyze an aresse [4]. A reversible Data Hiing metho base on wavelet sprea spectrum an histogram moification. Using sprea spectrum scheme ata is embee in the coefficients of the integer wavelet transform in high frequency bans [5].A lossless ata hiing metho for igital images using IWT an embeing base on threshol is one. Data are embee into the LSB planes of high frequency integer wavelet coefficients whose magnitue are lesser than a chosen threshol [6]. Data is embee in the bit planes of color component of the Integer wavelet transforme image. Bit plane complexity segmentation is use. To estimate the complexity a particular criteria is use an the IWT coefficient areas which can be replace to maintain imperceptibility is use[7].reversible ata Hiing Scheme for binary images is suggeste. JPEG2000 compresse ata is use an the bit-epth of the quantize coefficients are also embee in coe-blocks [8]. Manuscript receive November 5, 2009 Manuscript revise November 20, 2009
IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 251 2. Integer-To-Integer Wavelet Transforms In conventional wavelet transform reversibility is not achieve ue to the floating point wavelet coefficients, we get after transformation. When we take the inverse transform the original pixel values will get altere. When we transform an image block consisting of integervalue pixels into wavelet omain using a floating-point wavelet transform an the values of the wavelet coefficients are change uring watermark embeing, the corresponing watermarke image block will not have integer values. When we truncate the floating point values of the pixels, it may result in loss of information an reversibility is lost. The original image cannot be reconstructe from the watermarke image. In conventional metho wavelet transform is one as a floating-point transform followe by a truncation or rouning an it is impossible to represent transform coefficients accurately. Information will be potentially lost through forwar an inverse transforms. In view of the above problems, an invertible integer-tointeger wavelet transform base on lifting is use in the propose scheme. It maps integers to integers which are preserve in both forwar an reverse transforms. There is no loss of information.wavelet or subban ecomposition associate with finite length filters is obtaine by a finite number of primal an ual lifting followe by scaling.in the iscussion, we consier eightbit grayscale images an enote the least significant bitplaneas the 1st bit-plane, the most significant bit-plane the 8th bit-plane. In the commonly use grayscale images the stuy shows binary 0s an 1s are almost equally istribute in the lower bit-planes. The bias between 0s an 1s starts graually increasing in the higher bit-planes. This kin of bias inicates reunancy, implying that we can compress bits in a particular bit-plane or more than one bit-plane to leave space to hie other ata like text or image as watermark. Image transforms offer a larger bias between 0s an 1s in the wavelet omain than in the spatial omain. To eliminate more reunancy to embe ata an to avoi roun-off error, we propose to use the secon generation wavelet transform such as IDWT which maps integer to integer. This technique is base on the lifting scheme. 2.1 Bit-plane Embeing Using Arithmetic Coing Stuy has reveale that bias between binary 0s an 1s starting from the 2 n bit- plane of the IDWT coefficients increases than in the spatial omain. The higher the bitplane, the larger the bias. But alterations mae in higher bit-plane will lea to egraation of image quality. In orer to have the watermarke image perceptually the same as the original image, we choose to hie ata in one or more mile bit planes in the IDWT omain. The approximate coefficients in the LL subban contribute to visual perception. So specifically the LH, HL an HH subbans are use for watermark embeing. In the chosen bit-plane of the mile an high frequency subbans, the arithmetic coing is use to losslessly compress binary 0s an 1s because of its high coing efficiency. 3. Propose Scheme: The given image is ecompose into its frequency components using suitable wavelet transform. We have use the integer iscrete wavelet transform IDWT an the pixel values are transforme in the forwar an reverse irections losslessly. In the propose scheme the watermarke bits are embee into bit planes. The original image is preprocesse by performing lifting scheme. Now integer to integer wavelet transform is performe to ecompose the image into its components namely, Approximate coefficients, horizontal, vertical coefficients an iagonal coefficients. We use the horizontal vertical as well as the iagonal etaile bans to embe the watermark. We chose a bit plane of the etaile bans. The original bits in the selecte plane are compresse losslessely to create space for embeing the payloa bits. The compression exploits the fact that 0 san 1 s are nonuniformly istribute as we move from least significant bit plane to higher ones After compression necessary heaers are generate reflecting the original bit istribution in the chosen plane of the quarants. 3.1Embeing Process: For a given image of size M x N in which the gray scale set {1,2..255} inicate the pixel values an the wavelet coefficients are represente using eight bits. All the LSBs in a block represent the lowest bit plane, the next significant bits form the next plane an so on till the most significant bits form the most significant plane. Watermark bits are embee in the chosen bit plane. Let B represent original bits in the chosen plane an CB the compresse bits. Let W be the watermark bits. 16 Bits 16 Bits CH Heaer CV Heaer 16 Bits CD Hea er 16 Bits CH Length 16 Bits CV Leng th 16 Bits 32 Bits CD Length Watermar k Length CH, CV, CD heaers represents the bit istribution neee for arithmetic encoer an ecoer use for compression. CH, CV an CD Length represent the length of compresse bit stream in the chosen plane of the LH, HL an HH components. Bit Plane Ientification shows
252 IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 [ 8 th 7 th 6 th 5 th 4 th 3 r 2 n 1 st ] are the plane ientifiers.1 st plane represents the least significant plane an 8 th plane represents the most significant plane. Origin al Water marke Integer Wavele t Decom Approximat e Invers e IWT 3.2Embeing Algorithm Embee H,V & D bits Componen Watermar Fig 1. Embeing Process 1. Rea the original Image an ecompose it into its sub bans. 2. Separate the H, V an D etaile bans for watermarking. 3. Construct binary images of H,V an D of the chosen bit plane. 4. Compress the original bits in the chosen plane of these bans an erive the necessary heaers neee for the arithmetic encoer an ecoer. 5. Rea the water mark an convert it in to a bit string. 6. Now concatenate the heaer length, heaer, compresse bit stream CH, CV, CD an the watermark bits to a single bit stream. 7. Start embeing bit stream in to the bit plane of H. If not over continue in V an then in D an get the marke components of the image. 8. Now compute the inverse integer wavelet transform of the watermarke image from A an the embee H,Van D components to get the watermarke image. 3.3Extraction Algorithm Select H, V & D Comp Heaer Informati Embein g Compresse Original Choose Bit l ( ) Compressi on 1. Rea the watermarke image an take the integer wavelet transform to get the embee H, V an D sub bans an the unmarke approximate coefficients. 2. Separate the heaer, compresse H, V an D sub bans an the watermark bits. 3. Remove the watermark bits an ecompress the planes of the H, V an D sub bans to get the reconstructe sub bans. 4. Take inverse integer transform of the reconstructe H, V an D sub bans after Water marke Origin al I ecompression along with the unmarke approximate component to get the original image. 4. Experimental Results an Discussions: 4.1 Watermarke Image Quality Performance measure Watermarking the original image slightly egraes the original images as far as peak signal to noise ratio (PSNR) is concerne. But it is well within the visual perception an we o not reaily visualize the watermark an the egraation. The visual quality of the marke image is measure in PSNR. The mean square error (MSE) Table 1 Image Quality Teste for ifferent Gray Scale Images for each Payloa using bior 3.3 wavelet Watermark Image Size Integer Wavele t Decom Invers e IWT Approximat e Payloa bpp Select H, V & D Comp Decompre ssion Heaer Informati Separate Chosen Bit Separate original Original Compress H,V & D e Componen Watermar Fig 2. Extraction Process Lena Baboon Barbara 10 0.0003 32.81 28.61 31.28 50 0.01 32.78 28.60 31.22 100 0.04 32.67 28.57 31.10 150 0.09 32.46 28.53 30.74 200 0.15 32.10 28.47 30.29 250 0.24 31.69 x 31.01 300 0.34 31.28 x 29.94 350 0.47 31.18 x x 400 0.61 x x x 450 0.77 x x x
IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 253 34 Lena 33 Baboon Image Q uality (PSNR B) 32 31 30 29 Barbara c 28 0 0.1 0.2 0.3 0.4 0.5 Payloa bpp Fig 3 Comparison of embeing Capacity in bpp versus istortion in PSNR for ifferent Grayscale Images Image Quality (PSNR B) Table 2 Embeing Capacity of bit plane 4 an bit plane 5 Embee Bits PSNR Bit Plane 5 Bit Plane 4 100 34.50 35.18 2500 34.45 35.14 22500 33.92 34.87 40000 34.86 34.72 50625 33.83 34.64 62500 33.07 34.59 90000 32.36 x 122500 32.13 x 36 35 34 33 32 Bit Plane 5 Bit Plane 4 31 0 50000 100000 150000 Embee Bits Fig 4 Comparison of embeing Capacity an Image Quality in ifferent bit planes. a b e Fig 5 Original Image an Watermarke Image. a. Original Image, b. 30.01 B with 129600 bits c. 30.36B with 105625 bits,. 30.62 B with 78400 bits, e. 30.93 B with 55225 bits, f. 31.36 B with 22500 bits inicate the ifference between the original image an the watermarke image. 2 255 PSNR = 10log10 (1) MSE 1 n ( () ()) 2 ' MSE = I i I i (2) n i= 1 Where I an I are the original an watermarke Images respectively, n is the total number of pixels. 255 refer to the maximum possible pixel value in an eight bit image. Higher PSNR represents better signal quality. Table 1 shows the Image Quality of ifferent Gray scale images for each payloa. The embeing capacity is image epenent an is also base on the bit istribution of the chosen bit plane. The table shows Lena has better embeing capacity than Baboon an Barbara. Figure 4 inicates the comparison of the images for ifferent payloas. Table 2 shows the embeing capacity of lower bit planes is lesser than the higher bit planes. Experiment is conucte on bit plane 4 an bit plane 5.results show bit plane 5 has more embeing capacity but since it is more significant plane, the PSNR is slightly lesser in this plane, than in Plane 4. Figure 5 inicates these results. Table 3 shows the performance of ifferent wavelets on Lena, Baboon an Barbara images. The PSNR for the payloa of 10000 bits is shown along with the mean square error between the original an the water marke image. f
254 IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 Table 3. Performance of various wavelets an their Image Quality in PSNR for a fixe payloa of 10000 bits Wavel et Type X inicates capacity is insufficient for embeing. a Images Lena Baboon Barbara PSNR MSE PSNR MSE PSNR MSE coif1 30.81 51.41 29.23 72.27 30.64 59.61 9.7 34.81 20.14 29.70 64.27 32.06 37.28 cf 2.2 34.92 20.09 29.35 67039 31.83 41.61 b2 31.77 41.52 29.40 63.37 30.85 50.91 sym 2 31.77 41.5 29.40 63.37 30.86 50.91 bior 1.1 bior 3.3 bior 6.8 rbio 1.1 rbio 3.3 rbio 6.8 32.56 34.09 29.26 64.78 30.88 51.08 32.87 33.48 28.61 82.67 31.257 46.79 X X 30.25 67.67 32.475 41.52 X X 30.01 54.68 31.873 39.31 X X 28.65 78.679 31.125 47.84 34.50 22.162 29.37 71.65 31.98 40.53 c Fig 6. Original an watermarke images: (a) Original Image (b) Watermarke with 10,000 bits at-34.92 B (c) 62,500 bits error 16 PSNR 33.4623 ()122500 bits error 16 PSNR 32.4736 b Image Quality (PSNR B) 36 35 34 33 32 31 Coif 1 cf 2.2 9.7 bior 3.3 rbio 6.8 sym 2 30 0 0.2 0.4 0.6 0.8 Payloa (bpp) Fig 7. Performance of ifferent Wavelets teste on Lena Image for Image Quality (PSNR) vs Payloa (bpp). 5. Conclusion Lossless image watermarking is one an is completely reversible. Arithmetic coing use for compression guarantees complete reversibility. Lower bit planes have lower embeing capacity but since they are less significant for visual perception image quality is better than in higher bit planes. Performance of various wavelet families are shown. References [1] J. Fririch, M. Goljan an R. Du, Invertible authentication, Proc. SPIE, Security an Watermarking of Multimeia Contents, pp. 197-208, San Jose, CA, January (2001). [2] Goljan, M., Fririch, J., Du, R., "Distortion-Free Data Embeing for Images ", 4 th Information Hiing Workshop, Pittsburgh, Pennsylvania, April, 2001 [3] J. Tian: Reversible Data Embeing Using a Difference Expansion. IEEE Transactions on Circuits an Systems for Vieo Technology, Aug. 2003, 890-896. [4] Y. Q. Shi, Reversible ata hiing, Proceeings of International Workshop on Digital Watermarking, Seoul,Korea, Oct. 1 to Nov. 2, 2004. [5] G. Xuan, Y. Q. Shi, Z. Ni, Lossless ata hiing using integer wavelet transform an sprea spectrum, IEEE International Workshop on Multimeia Signal Processing, Siena, Italy, September 2004 [6] G. Xuan, Y. Q. Shi, C. Yang, Y. Zheng, D. Zou, P. Chai,: Lossless ata hiing using integer wavelet transform an threshol embeing technique. IEEE International Conference on Multimeia an Expo (ICME05), Amsteram, Netherlans, July, 2005. [7] IJCSNS International journal of Computer Science an Network Security.Vol 7 No 7 July 2007 Steganography Using BPCS To The Integer Wavelet Transforme Image K.Ramani,Dr.E.V.Prasa,Dr.S.Varaarajan [8] Lossless ata hiing using bit epth embeing for JPEG2000 compresse bit-stream. Feb 2009,Volume 6, No.2(serial No.51), Journal of Communication an Computer,Shogo Ohyama, Michiharu Niimi,Kazumi Yamawaki,Hieki Noa.
IJCSNS International Journal of Computer Science an Network Security, VOL.9 No.11, November 2009 255 S. Kurshi Jinna Complete her B.E in Electronics an Communication Engineering from Thiagarajar College of Engineering, Maurai, in 1985 an M.E(Hons) in Computer Engineering from VJTI, University of Mumbai an oing Ph.D in faculty of information an communication in Anna University, Chennai. She is currently working as Professor & hea of the epartment, Computer Science an Engineering in PET Engineering College, Vallioor, Inia. Dr. L.Ganesan complete his B.E in Electronics an Communication Engineering from Thiagarajar College of Engineering, Maurai an M.E in Computer Science an Engineering from Government College of Technology, Coimbatore. He complete his Ph.D from Inian Institute of Technology, Kharagpur in the area image processing. He has authore more than fifty publications in repute International Journals. His area of interest inclues image processing, multimeia an compressions. He is currently working as hea of the epartment of Computer science an engineering, A.C. College of Engg. An Technology, Karaikui, Inia