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 2287-9137 (Online) ISSN 1226-7953 (Print) HEVC a), b), a) H.265/HEVC Video Watermarking Method with High Image Quality Young-Ho Seo a), Yong-Seok Lee b), and Dong-Wook Kim a). HEVC Random Access B I P I P I, P, B., B. Abstract In the field of video watermarking, which adds a time axis to the watermarking of the existing still image, it is important to embed a watermark at a level that can claim copyright while minimizing the overall video quality. In this paper, we focus on the fact that B slices refer to blocks of I and P slices in the Random Access compression method of HEVC. Therefore, We propose that watermark is embedded only in I and P slices and extracted in I, P, and B slices. This can minimize the decreases of the overall image quality of the image and protect the copyright of the B slice without directly embedding the watermark. Keyword : HEVC, DWT, Watermark, Blind, Robust a) (Department of Electronic Materials Engineering, Kwangwoon University) b) (Korea Electronics Technology Institute) Corresponding Author : (Dong-Wook Kim) E-mail: dwkim@kw.ac.kr Tel: +82-2-940-5167 ORCID:https://orcid.org/0000-0002-4668-743X This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education(NRF-2016R1D1A1B03930691). Manuscript received September 10, 2018; Revised December 17, 2018; Accepted December 18, 2018.... (robust) (fragile). Copyright 2016 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 (http://creativecommons.org/licenses/by-nc-nd/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited and not altered.
(JBE Vol. 24, No. 1, January 2019), [1][2]. 2, [1].. I (slice). I (intra), I B P., [3]... [4]. Gaj I 4x4 (transform block) [5]. I LowDelay GOP4. 1,280 720 HEVC.,, (location map). 100, QP16. Alenizi FullHD 1,920 1,080, LowDelay GOP4, 99 QP20 [6].. 1024, HEVC All Intra, Low-delay, Random Access, HEVC.. 2 3. 4.. Random Access B,. B I P. B. HEVC. B B. 48 50fps(frame per second) I 1 P 6 49. I P HEVC. HEVC. DWT(discrete wavelet transform), DWT QIM [3].
2: HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality),.. 1. 1. I P Y. CbCr No Original RGB video I or P slice? Yes Convert to YCbCr format Y channel n-level 2DDWT Deciding quantization step size Multiple WM embedding n-level inverse DWT Convert to RGB format Video sequencing Watermarked image 1. Fig. 1. Watermark embedding procedure Watermark data Data scrambling Key I P RGB YCbCr 2 DWT (2DDWT). DWT n. n 2DDWT 4. 16. (S = LL, LH, HL, HH). 2DDWT. QIM. LH, HL, HH LL. (1). if if if (1). (2). for for (2) i,.. DWT 16.. 2. 2 (1).,. 2 0 1.
(JBE Vol. 24, No. 1, January 2019) c i ' 3a Sn 1. 2a Sn 0 Watermarked /attacked image -3a Sn -2a Sn a Sn -a Sn 0 1 a 0 Sn 2a Sn 3a Sn c i Convert to YCbCr format Y channel n-level 2DDWT 0 1 -a Sn -2a Sn Deciding size of quantization step Multiple WM extraction 2. Fig. 2. Quantizer to embed watermark bit De-scrambling Single WM formation Key 2 (3). (3) i,. if if if if if if 2DDWT. 2.. 3,. B I P. Extracted WM 3. Fig. 3. Watermark extraction procedure 3. RGB YCbCr 2DDWT., 2DDWT. (2).,. n 2DDWT i (4). mod 16...
2: HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality) III.. C/C++, PC Intel Core i7-2700k CPU @ 3.50GHz, 16GB RAM, 64 Windows 7 Ultimate K. JCT [7]. 1. 1. Table 1. Parameter values for each subband Subband LL LH HL HH 1.5 2.0 2.0 2.0 2.5 8.0 8.0 8.0 0.04 0.2 0.2 0.28 4 2.,, 2DDWT n (1). 4. Fig. 4. Watermark data 2. PSNR, NCC. 43dB.,. 2DDWT. 40/100 JPEG, 5 5,,, 3%... 2. : (PSNR[dB]) NCC Table 2. Experimental results for each kind of images: image quality after watermark embedding (PSNR[dB]) and NCC value of extracted watermark Attacks Robustness PSNR NCC Only watermarked 42.94 1.00 100 46.21 1.00 JPEG quality to 100 80 36.30 1.00 60 33.92 0.98 40 32.31 0.88 0.005 50.19 1.00 Gaussian noise 0.01 39.01 1.00 0.02 34.27 0.96 Salt&pepper noise 1% 41.43 1.00 3 3 34.14 0.99 Median filtering 5 5 30.32 0.91 7 7 28.48 0.79 Average filtering 3 3 32.93 0.98 5 5 29.34 0.87 Gaussian filtering 3 3 35.07 1.00 Sharpening 34.28 1.00 Histogram equalization 15.62 0.60 Contrast (-20) 27.18 1.00 2 47.14 1.00 1.8 47.35 1.00 Scaling to 1.5 47.20 1.00 0.8 39.93 1.00 0.5 34.02 1.00 0.25 29.12 0.85 Cropping 25% 12.71 0.81 90 19.33 0.77 Rotation 60 17.49 0.85 45 13.86 0.88 30 14.51 0.92 Rotation 0.1 26.53 0.91 (unrestored) 0.2 25.68 0.74 Average 31.99 0.89
102 방송공학회논문지 제24권 제1호, 2019년 1월 (JBE Vol. 24, No. 1, January 2019) 공격에 사용된 영상은 Full HD(1,920 1,080) 해 상도의 500 프레임 50fps 영상을 이용하여 실험하였다. 50fps영상의 intra period는 48이다. I 및 P 프레임에만 워터 마크를 삽입하고 QP22, 27, 32 및 37로 인코딩과 디코딩을 거친 영상에서 워터마크를 추출하였다. 그림 5에 워터마크 를 삽입한 이후에 HEVC 인코딩과 디코딩을 거친 결과 영 상을 나타냈다. 앞서 설명한 것과 같이 HEVC의 low-delay 모드로 압축할 경우에 I 슬라이스에만 워터마크 데이터를 삽입하였다. 하지만 I 슬라이스 뿐만 아니라 P 슬라이스에 서도 워터마크를 추출하였다. HEVC (a) (b) (a) (b) (c) (d) (e) (f) 그림 6. 워터마크가 삽입된 이후에 (QP=22, HEVC 인코딩) HEVC 디코딩 결과 영상 (a) 0, (b) 2, (c) 4, (d) 6, (e) 8, (f) 10번째 프레임 Fig. 6. HEVC decoding result after watermark embedding (QP=22, HEVC Encoding) (a) 0th, (b) 2nd, (c) 4th, (d) 6th, (e) 8th, (f) 10th frame (c) (d) 표 3. HEVC 압축 공격에 대한 PSNR 결과 Table 3. PSNR results for HEVC compression attack QP (e) (f) 그림 5. 워터마크가 삽입된 후에 HEVC 공격 결과 (a) 원본, (b) QP=1, (c) QP=22, (d) QP=27, (e) QP=32, (f) QP=37 Fig. 5. HEVC attack result after watermark embedding (a) original, attacks with (b) QP=1, (c) QP=22, (d) QP=27, (e) QP=32, (f) QP=37 워터마크 삽입한 영상 압축 후 화질 열화 정도를 표 3에 나타내었다. 워터마크를 삽입하지 않은 원본 영상의 압축 후 화질 열화정도는 QP별로 인코딩 디코딩한 영상을 압축 하지 않은 영상과 PSNR 수치로 나타내었다. 워터마크 삽 입 후 영상의 압축 후 화질 열화정도는 QP별로 인코딩 디 코딩한 영상을 압축하지 않은 영상과 PSNR 수치로 나타내 었다. 표 3에서 알 수 있듯이 워터마크를 삽입했음에도 압 축 후 화질 저하가 적다는 것을 알 수 있다. 22 27 32 37 Compression Ratio 142:1 428:1 922:1 1817:1 Original PSNR[dB] 32.83 31.70 30.50 29.22 Watermarked PSNR[dB] 32.77 31.64 30.42 29.14 그림 7에서는 한 GOP 내에서 I, P, B 슬라이스 별 워터마 크 추출율을 보여주고 있다. 이는 I와 P슬라이스에만 워터 마크를 삽입 했지만 B슬라이스에서도 워터마크가 검출되 며, QP가 높으면 B슬라이스 검출율이 높을 가능성이 크기 도 하다는 것을 보여준다. 그림 8은 하나의 GOP 내의 9개 프레임을 종합하여 워터 마크를 추출한 결과이다. 이는 각 슬라이스에서 추출된 워 터마크를 통계적으로 종합하여 하나의 워터마크로 만드는 방법으로, 압축률 922:1에 해당하는 QP32에서도 NC값 0.9182의 높은 추출율을 보이고 있다. 그림 9에는 추출된 워터마크를 나타냈다.
2: HEVC (Young-Ho Seo et al.: H.265/HEVC Video Watermarking Method with High Image Quality) Gaj [5]. Gaj QP QP. Basketball. Gaj 20, 24, 28, 32 36 QP, HEVC 0.8199, 0.7396, 0.7189, 0.7071 0.6715 NC. 22, 27, 32 37 QP, 1, 0.97, 0.63 0.07 NC., 32 QP Gaj. (a) (b) (c) (d) 9. (a) I, (b) B, (c) P, (d) Fig. 9. Example of watermarks extracted from each slice, (a) I slice, (b) B slice, (c) P slice, (d) Statistically summarized final watermark IV. Intra, Random Access..,.. (References) 7. Fig. 7. Ratio of extracted watermark 8. GOP Fig. 8. Extracted final watermark from one GOP group [1] W. M. Chen, C. J. Lai, H. C. Wang, H. C. Chao and C. H. Lo, H.264 video watermarking with secret image sharing, 2009 IEEE International Symposium on Broadband Multimedia Systems and Broadcasting, Bilbao, Spain, pp.1-7, 2009, doi:10.1109/isbmsb.2009.5133838. [2] M. Fallahpour, S. Shirmohammadi, M. Semsarzadeh and J. Zhao, Tampering Detection in Compressed Digital Video Using Watermarking, IEEE Transactions on Instrumentation and Measurement, Vol.63, No.5, pp.1057-1072, May 2014, doi:10.1109/tim.2014. 2299371. [3] K. Ogawa and G. Ohtake, Watermarking for HEVC/H.265 Stream, 2015 IEEE International Conference on Consumer Electronics (ICCE), Las Vegas, USA, pp.102-103, 2015, doi:10.1109/icce.2015. 7066337. [4] D-W. Kim, Y-G.Choi, H-S. Kim, J-S. Yoo, H-J. Choi and Y-H. Seo, "The problems in digital watermarking into intra-frames of H.264/ AVC", Image and Vision Computing, Vol.28, No.8, pp.1220-1228, May. 2010, https://doi.org/10.1016/j.imavis.2009.12.006 [5] S. Gaj and P. K. Bora, A robust watermarking scheme against re-compression attack for H.265/HEVC, IEEE 5th National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics (NCVPRIPG), Patna, India, pp.1-4, Dec. 2015, doi:10.1109/ NCVPRIPG.2015.7490065.
(JBE Vol. 24, No. 1, January 2019) [6] F. Alenizi, F. Kurdahi, A. Eltawil, and A. Aljumah, DWT-based watermarking technique for video authentication, 2015 IEEE International Conference on Electronics, Circuits, and Systems (ICECS), Cariro, Egypt, pp.41-44, Dec. 2015, doi:10.1109/icecs.2015.7440244 [7] Vidio set of JCT, ftp://hevc@ftp.tnt.uni-hannover.de/testsequences/ (accessed Jan. 15, 2019). - 1999 2 : () - 2001 2 : () - 2004 8 : () - 2005 9 ~ 2008 2 : - 2008 3 ~ : - ORCID : http://orcid.org/0000-0003-1046-395x - :, 2D/3D,, SoC - 2015 2 : () - 2017 2 : () - 2017 3 ~ : (KETI) - :. 2D/3D - 1983 2 : () - 1985 2 : - 1991 9 : Georgia () - 1992 3 ~ : - ORCID : http://orcid.org/0000-0002-4668-743x - : 3D,, VLSI Testability, VLSI CAD, DSP, Wireless Communication