3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) (Special Paper) 22 3, (JBE Vol. 22, No.

Size: px

Start display at page:

Download "3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) (Special Paper) 22 3, (JBE Vol. 22, No."

인선 비
5 years ago
Views:

1 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) (Special Paper) 22 3, (JBE Vol. 22, No. 3, May 2017) ISSN (Online) ISSN (Print) DWT a), a), b), a) An Empirical Digital Image Watermarking using Frequency Properties of DWT I-Seul Kang a), Yong-Seok Lee a), Young-Ho Seo b), and Dong-Wook Kim a).,. 2 (2-Dimensional Discrete Wavelet Transform, 2D-DWT),. (Quantization Index Modulation, QIM),...,. Abstract Digital video content is the most information-intensive and high-value content. Therefore, it is necessary to protect the intellectual property rights of these contents, and this paper also proposes a watermarking method of digital image for this purpose. The proposed method uses the frequency characteristics of 2-Dimensional Discrete Wavelet Transform (2D-DWT) for digital images and digital watermark on global data without using local or specific data of the image for watermark embedding. The method to insert digital watermark data uses a simple Quantization Index Modulation (QIM) and a multiple watermarking method that inserts the same watermark data in multiple. When extracting a watermark, multiple watermarks are extracted and the final watermark data is determined by a simple statistical method. This method is an empirical method for experimentally determining the parameters in the watermark embedding process. The proposed method performs experiments on various images against various attacks and shows the superiority of the proposed method by comparing the performance with the representative existing methods. Keyword : digital image watermarking, frequency property, 2-dimensional discrete wavelet transform (2D-DWT), global data, empirical method Copyright 2017 Korean Institute of Broadcast and Media Engineers. All rights reserved. This is an Open-Access article distributed under the terms of the Creative Commons BY-NC-ND ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited and not altered.

2 (JBE Vol. 22, No. 3, May 2017).. (ownership) (intellec- tual property right), 30 [1]. [2]. (digital watermarking, DWM) [1],. DWM (proof of owner), (au- thentication), (integrity), / (copy control), (tracking), (broadcast- ing monitoring) [1],,. DWM (watermark, WM) blind non-blind [1]. Non-blind WM, blind. blind Non-blind [1]. WM a) (Department of Electronic Materials Engineering, Kwangwoon University) b) (School of Liberal Arts, Kwangwoon University) Corresponding Author : (Dong-Wook Kim) dwkim@kw.ac.kr Tel: ORCID: () (NRF-2016R1D1A1B ). Manuscript received February 28, 2017; Revised May 22, 2017; Accepted May 22, [1]. DWM WM (robustness) fragile [1]. WM (, invisible) WM (, visible) [1]. WM blind. DWM WM WM. 2 (2-dimensional discrete wavelet transform, 2D-DWT),. 2D- DWT. 2D-DWT. 2D-DWT 1- (level) [3~6], LL1(,, L, H.. LL1-1 (subband) ) 10 DWM, JPEG Salt& pepper, LL1 WM. [7], HH1. 2-, [8] D-DWT HH WM. WM, average filtering, median filtering, crop

3 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT). 2D-DWT DCT(discrete cosine transform) [9], HL1, LH1, HL2, LH2, HH2 8 8 DCT (6,7) (7,6) WM. median HH1 HH WM WM [10]. (color) DWM [11], (human visual system, HVS) blue WM 0.9. [12] RGB YCbCr HVS Cb, 4-2D-DWT HL4 WM. 0.9 NCC rotation. WM QIM(quantization index modulation) [13~15], [13], [14][15]. [16]. 1-2D-DWT, LL1 4 4 SVD(singular value decomposition) WM. WM (non-blind)., WM. WM, WM WM WM. WM (trade-off). 2D-DWT WM, WM. 2D-DWT, WM WM Image I One level transform g[n] 1/2 h[n] 1/2 g[n] h[n] g[n] 1/2 1/2 1/2 LL1 To next level transform LH1 HL1 LL3 HL3 LH3 HH3 LH2 HL2 HH2 HL1 Horizontal transform h[n] 1/2 HH1 LH1 HH1 Horizontal transform Gain (a) Level 3 Level 2 Level 1 (b) f n /8 f n /4 f n /2 f n... Frequency (c) 1. 2D-DWT : (a), (b) 3-2D-DWT, (c) 3- DWT. Fig. 1. 2D-DWT and the frequency band of subbands: (a) filter bank, (b) subbands after 3-level 2D-DWT, (c) frequency band for 3-level DWT.

(JBE Vol. 22, No. 3, May 2017) WM.. 2 2D- DWT. 3 DWM. 4,.

1(a) 2D-DWT (filter bank), 2, ( ) ( ). 1/2 1/2. LL.

. 1- DWT 0, LL1 2 2 0 1 (HL1, LH1, HH1) (a) (b) (c) (d)

2D-DWT : (a) 1,024 1,024, (b) (a) 4-2D-DWT, (c) (a) 512

An example of 2d-DWT: (a) 1,024 1,024 original image, (b)

4 (JBE Vol. 22, No. 3, May 2017) WM.. 2 2D- DWT. 3 DWM. 4, D-DWT 2D-DWT 2 [17]. 1(a) 2D-DWT (filter bank), 2, ( ) ( ). 1/2 1/2. LL. 2(b) 3-. 1(c) 3-1 DWT.. 1- DWT 0, LL (HL1, LH1, HH1) (a) (b) (c) (d) (e) (f) 2. 2D-DWT : (a) 1,024 1,024, (b) (a) 4-2D-DWT, (c) (a) , (d) (b) 3-2D-DWT, (e) (b) -4, (f) (d) -3 Fig. 2. An example of 2d-DWT: (a) 1,024 1,024 original image, (b) result from 4-level 2D-DWT for (a), (c) down-sampled result from (a), (d) resu;t from 3-level 2D-DWT for (c), (e) four level-4 subbands of (b), (f) four level-3 subbands of (d).

5 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT).., VR(virtual reality) AR(augmented reality). (scene).,., 2D-DWT? 2. 1,024 1,024 St. George Church(a) 1/2 (c), 4- (b), 3-2D-DWT(d). 1(c) (c) (a). 2D-DWT (b) (d). 2 (b) -4 (e) (d) -3 (f). (e) (f).. 2(b), -1,.. DWT DWM., DWM. 1. WM. YCbCr Y DWM. 2D-DWT, WM. WM (bit), WM ( WM). WM QIM. QIM [18], ( ) WM WM. DWM., WM WM. 3. Fig. 3. Quantizer c i ' 2.5Δ L 1.5Δ L 0.5Δ L Δ L -1.5Δ L -2.5Δ L 0 1 c i

(JBE Vol. 22, No. 3, May 2017) (a) (b) (c) (d) 4. 1.024 1,024 : (a) St. George Church, (b) Two Horses, (c) Neuschwanstein Castle, (d) Four Kittens. Fig. 4. The 1.

512 512 2-2 WM, 3-3 4-4.,. DWM WM WM PSNR(peak signal-to-noise ratio) [19] 45[dB].. 2,. 1. Table 1.

1,024 1,024 5 Gaussian noise addition 1%, 2%, 3% 1,920 1,080 5 Sharpening 2, 3, 4 2,048 2,048 6 Rotation 15, 30, 45 2. Table 2.

6 (JBE Vol. 22, No. 3, May 2017) (a) (b) (c) (d) ,024 : (a) St. George Church, (b) Two Horses, (c) Neuschwanstein Castle, (d) Four Kittens. Fig. 4. The ,024 test images: (a) St. George Church, (b) Two Horses, (c) Neuschwanstein Castle, (d) Four Kittens., 1.,,. 1,024 1, D-DWT 2-1, 2- DWM WM, ,. DWM WM WM PSNR(peak signal-to-noise ratio) [19] 45[dB].. 2,. 1. Table 1. Summary of test experiments Resolution Highest 2D-DWT level Type Attacks Intensity JPEG compression 40, 60, Median filtering 3 3, 5 5, 7 7 1,024 1,024 5 Gaussian noise addition 1%, 2%, 3% 1,920 1,080 5 Sharpening 2, 3, 4 2,048 2,048 6 Rotation 15, 30, Table 2. Summary of the three experiments Experiment Final WM data Quantization step size : All the embedded WM sets One WM set by statistical method One WM set by statistical method Same for all subbands Same for all subbands Adjustment according to subband energy ~, DWM 45[dB]. WM

7 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) (a) (b) (c) (d) (e) (f) 5. 1,024 1,024 1 : (a) JPEG, (b), (c), (d), (e), (f). Fig. 5. Experiment 1 results of 1,024 1,024 images for: (a) JPEG compression attacks, (b) median filtering attacks, (c) Gaussian noise addition attacks, (d) sharpening attacks, (e) rotation attacks, (f) average for all the attacks. (a) (b) (c) (d) (e) (f) 6. 1,024 1,024 2 : (a) JPEG, (b), (c), (d), (e), (f). Fig. 6. Experiment 2 results of 1,024 1,024 images for: (a) JPEG compression attacks, (b) median filtering attacks, (c) Gaussian noise addition attacks, (d) sharpening attacks, (e) rotation attacks, (f) average for all the attacks.

8 (JBE Vol. 22, No. 3, May 2017) WM WM NCC(normalized cross correlation) [20]. 1,024 1,024 5,. 5, 2D-DWT, (f).. 2D-DWT : WM 1 WM WM. WM, ,024 1,024 6,. 1, 1,. JPEG,. 1 NCC 5. (f) : 1 2 WM.., WM,.. WM WM. 7,, 2D-DWT. 2D-DWT LL. LL 0 ( ), 0. (1) LL, RMS(root-mean-square). f or f or Image + - n-level 2DDWT Subband energy calculation Attack 7. Fig. 7. Experiment to estimate the effect of an attack WM 7 2 1,024 1, D-DWT 3. 6(f) 1,024 1, D-DWT. 3 (a) 7, (b) 2.

9 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) 3. 1,024 1, D-DWT Table 3. Example result from experiment on the effect of attack for 4-level 2D-DWT of 1,024 1,024 images. Attack (a) Effect of attack by energy (b) WM extraction rate by NCC LL HL LH HH LL HL LH HH No attack (original image) (71.624) pxl Sharpening 2 pxl pxl Gaussian 1% noise 2% addition 3% Median filtering JPEG compression quality Rotation / / / º º º ,. (a),.,., 4-.,, JPEG -4. (b).,.,,, WM. WM DWM. DWM 3., (a) (b), HL LH,. (2) (weighting factor, ) (1) ( ), LL, HL=LH, HH (3)

10 (JBE Vol. 22, No. 3, May 2017),. (3) WM (3). E Sn T Sn,h T Sn,l E Sn =e Sn (2). (threshold value) (4) (1),. 8. if if if 8. Fig. 8. Quantization range T Sn,l T Sn,h 2, 1 2 1,024 1, , JPEG 2. JPEG 2.,., e Sn (a) (b) (c) (d) (e) (f) 9. 1,024 1,024 3 : (a) JPEG, (b), (c), (d), (e), (f). Fig. 9. Experiment 3 results of 1,024 1,024 images for: (a) JPEG compression attacks, (b) median filtering attacks, (c) Gaussian noise addition attacks, (d) sharpening attacks, (e) rotation attacks, (f) average for all the attacks.

11 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT)..., 2D-DWT 4,096 16, Fig. 10. Results from Experiment 3 for all resolutions , (5)., 2D-DWT DWM 11(a), 11(a) WM 11(b). 2.1 Y RGB YCbCr. Y (5) 2D-DWT. (1), Original RGB image Watermark data Watermarked and attacked image CbCr Convert to YCbCr format Y Determine 2D-DWT level and transform Weighting factor calculation Convert to YCbCr format Y Determine 2D-DWT level and transform Weighting factor calculation Multiple WM embedding Data scrambling Multiple WM extraction Inverse 2D-DWT Key De-scrambling Key Convert to RGB format Single WM formation Watermarked image Extracted WM (a) (b) 11. DWM : (a) WM, (b) WM. Fig. 11. The proposed DWM scheme: (a) WM embedding procedure, (b) WM extracting procedure.

12 (JBE Vol. 22, No. 3, May 2017) (2) (4) (3). 3 WM. WM, [18] (linear feedback shift register, LFSR) [18][20] WM Exclusive-OR. (key) LFSR. WM, 2D-DWT Y, CbCr RGB. 2.2 DWM WM 11(b)., 2D- DWT, WM. WM, 3 WM. WM WM,. WM WM LFSR, WM Exclusive-OR WM. IV. 3 WM. 1. i7-2700k CPU 16GB RAM PC 64- Windows 7 Ultimate C/C++. WM 12 2, (M:N). DWM, [21~23]. 2D-DWT. WM D-DWT 4 16 WM. 4 (symmetric extension) [24]. 1,920 1, ,920 1,088. WM. 12. WM Fig. 12. Used WM 4. Table 4. Used images in experiments M:N Resolution DWT level # of images 1:1 4:3 16: ,024 1, , ,920 1, ,840 2, ,,,. 5 (pixel- value change attack), (geometric attack)

13 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT). 5. Table 5. Experimented attacks Category Type Intensity Pixel-value change attack Geometrical attack JPEG compression 80, 60, 40, 20/100 Gaussian blurring 3 3, 5 5 Average blurring 3 3 Median filtering 3 3 Sharpening - Gaussian noise addition 3% Salt&pepper noise addition 3% Histogram equalization - Contrast -20 Shrink 0,8, 0.5, 0.25 Magnify 2 Rotation π/6, π/4, π/3, π/2 Cropping 25% 2. 3 (, ) ( ) 1. 6, LL,.. 6. Table 6. Determined parameter values by experiment Parameter LL LH HL HH TSn,l TSn,h γsn WM PSNR (a) (b) (c) (d) (e) (f) 13. : (a) JPEG, (b),,, (c),,, Salt&pepper,, (d) (, ), (e), cropping, (f). Fig. 13. Experimental results for attacks: (a) JPEG compression, (b) Gaussian blurring, average blurring, median filtering, (c) sharpening, histogram equalization, Gaussian noise addition, Salt&pepper noise addition, contrast adjustment, (d) scaling (shrink, magnify), (e) rotation, cropping, (f) averages.

14 (JBE Vol. 22, No. 3, May 2017). WM, 40[dB], 39.67[dB]. 7. WM Table 7. Invisibility of the embedded WM Resolution PSNR[dB] ,024 1, , ,920 1, ,840 2, Average , ( ). PSNR ( P) WM NCC( N). PSNR, NCC, PSNR, NCC, PSNR NCC., (a), (b), (c), (d) (e), (f)., (a) JPEG, (b) (G blur), (Avg blur), (Med flt), (c), (sharpening), (Hist. Eq), (G noise), Salt&pepper (S noise), (Constrast). (d) (Shrink) (magnify), (e) (Rot) cropping(crop). (f) (Pixel Avg), (Geo Avg), (Total Avg)., (a) JPEG PSNR PSNR., JPEG quality 40/100 1,024 1,024 NCC. quality 40/100. (b) PSNR NCC ,024 1, NCC,. (c) PSNR NCC., PSNR, 0.6 NCC.. 4/5. (d) (e). (d) 1/4 1 NCC. 1/4 1,024 1, NCC. (e) π/2. 15[dB]. 0.9 NCC. π/3 π/4 1,920 1,080

42 [14] 64 64 40.48 [15] 64 64 39.88 [16] 32 32 44.08 14.

15 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT). (f)., 0.9 NCC. 4.. [8], [14], [15], [16],. WM. 8 WM DWM PSNR WM 4~16. PSNR,, 39.67[dB]. 8. Table 8. Comparison of Invisibility with existing schemes Scheme Amount of WM PSNR[dB] Ours ( )~ ( ) [8] [14] [15] [16] [8] [14] (a) (b) (c) (d) 14. : (a) [8][14], (b) [8][14], (c) [15], (d) [16]. Fig. 14. Robustness comparison with existing schemes: (a) pixel-value change attacks with [8] and[14], (b) pixel-value change attacks with [8] and [14], (c) pixel-value change attacks with [15], (d) geometric attacks with [16].

16 (JBE Vol. 22, No. 3, May 2017) WM WM BER(bit error rate), [15] [16] NCC. [8] [14] (a)(b), [15][16] (c)(d), (a) (c), (b) (d). 1 14,. [15], [16] 14(c) [15], 14(d) [16]. 14(a), [8] [14] BER. [14] 0.043, 0.050, [14] BER. BER, [8] 0.137, [14] 0.074, BER. (b) [8][14], BER. (Total Avg) [8] 0.039, [14] 0.075, [15], (c) NCC. JPEG 60/100, NCC. (d) [16] π/3, π/4, π/6 [16], 0.99 NCC. [16] 0.954, ,,.. 2D-DWT 2D-DWT,. WM WM. WM QIM,. WM WM WM, WM WM.,.,.. DWT, WM. (References) [1] I. J. Cox, et al., Digital watermarking and steganigraphy, Morgan Kaufmann Publisher, [2] Wlliam Stalling, Cryptography and network security, Prentice-Hall, [3] N. M. Makbol and B. E. Khoo, Robust blind image watermarking scheme based on redundant discrete wavelet transform and singular value decomposition, J. of Electronics and Communications, Vol. 67,

강이슬 외 인 의 주파수 특성을 이용한 실험적 디지털 영상 워터마킹 311 3 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) No. 2, pp. 102-112, Feb. 2013. C.

3060-3063, Nov. 2010. [5] E. Ganic and A. M. Exkicioglu, Robust embedding of visual watermarks using discrete wavelet transform and singular value decomposition, J. of Electronic Imaging, Vol. 14, No.

Fathy, Robust watermarking scheme based on RDWT-SVD: embedding data in all subbands, Artificial Intelligence and Signal Processing Symposium, pp. 48-52, June 2011. [7] P. P. Thulasidharan and M. S. Nair, QR code based blind digital image watermarking with attack detection code, J.

17 강이슬 외 인 의 주파수 특성을 이용한 실험적 디지털 영상 워터마킹 : DWT (I-Seul Kang et al.: An Empirical Digital Image Watermarking using Frequency Properties of DWT) No. 2, pp , Feb C. C. Lai and C. C. Tsai, Digital image watermarking using discrete wavelet transform and singluar value decomposition, IEEE Trans. on Instrumentation and Measurement, Vol. 59, No. 11, pp , Nov [5] E. Ganic and A. M. Exkicioglu, Robust embedding of visual watermarks using discrete wavelet transform and singular value decomposition, J. of Electronic Imaging, Vol. 14, No. 4, (043004), Dec [6] S. Lagzian, M. Soryani, and M. Fathy, Robust watermarking scheme based on RDWT-SVD: embedding data in all subbands, Artificial Intelligence and Signal Processing Symposium, pp , June [7] P. P. Thulasidharan and M. S. Nair, QR code based blind digital image watermarking with attack detection code, J. of Electronics and Communications, Vol. 69, No. 7, pp , Jul [8] R. Mehta, V. P. Vishwakarma and N. Rajpal, Lagrangian support vector regression based image watermarking in wavelet domain, Signal Processing and Integrated Networks, pp , Feb [9] J. Maedeh, S. Shadrokh and K. Nader, Robust image watermarking by multi resolution embedding in wavelet transform coefficients, ICEE 2015, pp , May [10] T. H. Nguyen, D. M. Duong and D. A. Duong, Robust and high capacity watermarking for image based on DWT-SVD, IEEE Int l Conf. on RIVF, pp , Jan [11] J. George, S. Varma and M. Chatterjee, Color image watermarking using DWT-SVD and Arnold transform, India Conference (INDICON), pp. 1-6, Dec [12] A. Roy, A. K. Maiti and K. Ghosh, A perception based color image adaptive watermarking scheme in YcbCr space, SPIN, pp , [4] 강이슬 Feb [13] B. Liao and J. Lv, A Novel watermark embedding scheme using compressive sensing in wavelet domain, The Open Cybernetics & Systemics Journal, Vol. 9, No. 1, pp. 1-6, Jun [14] H. T. Hu, Y. J. Chang and S. H. Chen, A progressive QIM to cope with SVD-based blind image watermarking in DWT domain, IEEE China SIP, pp , Jul [15] J. Ouyang, G. Coatrieux, B. Chen and H. Shu, Color image watermarking based on quaternion Fourier transform and improved uniform log-polar mapping, Computers & Electrical Engineering, org/ / j.compeleceng [16] Y. Xueyi, et al. "A SIFT-based DWT-SVD blind watermark method against geometrical attacks," IEEE, Intl. Cong. CISP, pp , Oct [17] R. M. Rao and A. S. Bopardikar, Wavelet transforms, AddisonWesley, [18] Y. S. Lee, Y. H. Seo, and D. W. Kim, Robust and Blind Watermarking for DIBR Using a Depth Variation Map, JBE Vol. 21, No. 6, pp , 2016 November. [19] R. C. Gonzales and R. E. Woods, Digital image processing, Pearson Prentice-hall, Upper Saddle River, NJ, [20] _register [21] [22] [23] VbgqFvntlBc [24] 저자소개 년 월 광운대학교 전자재료공학과 졸업 공학사) 년 월 현재 광운대학교 전자재료공학과 공학석사) 주관심분야 영상 신호처리 워터마킹 : ( ~ : ( - ORCID : : Virtual Reality,, 이용석 년 월 광운대학교 전자재료공학과 졸업 공학사) 년 월 광운대학교 전자재료공학과 졸업 공학석사) 주관심분야 영상 신호처리 워터마킹 설계 : ( : ( - ORCID : : 3D,, SoC

18 (JBE Vol. 22, No. 3, May 2017) : () : ( ) : ( ) ~ : ~ : - ORCID : - :, 2D/3D,, SoC : () : : Georgia ( ) ~ : ~ : - ORCID : - : 3D,, VLSI Testability, VLSI CAD, DSP, Wireless Communication

2 : HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality) (Regular Paper) 24 1, (JBE Vol. 24, No. 1, January

2 : HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality) (Regular Paper) 24 1, (JBE Vol. 24, No. 1, January 2: HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality) (Regular Paper) 24 1, 2019 1 (JBE Vol. 24, No. 1, January 2019) https://doi.org/10.5909/jbe.2019.24.1.97 ISSN