H.264/AVC a), a), b), b), a) A New Method for Thubnail Extraction in H.264/AVC Bitstreas Seung-Hwan Hong a), Jieon Ki a), Youngin Chin b), Jae-Cheol Kwon b), and Seoung-Jun Oh a) IPTV, DMB H.264/ AVC. H.264/AVC,. H.264/AVC. H.264/AVC. 8.66 db PSNR. Abstract Recently, thubnail techniques are required to index a high-perforance video at digital convergence-based ultiedia service like IPTV and DMB. Therefore a thubnail extraction ethod in H.264/AVC bitstreas has been proposed. However, thubnail quality deterioration proble at converting the general equation of spatial doain to frequency doain which is generated by not considering about H.264/AVC transfor and quantization processing and rounding-off operation in intra prediction. In this paper, we propose a new thubnail extraction ethod in H.264/AVC bitstreas. The proposed schee is based on H.264/AVC core-transfor for a thubnail extraction in frequency doain, and probability theory, intra rounding-off error copensation. Through the ipleentation and perforance evaluation, the subjective quality difference between the output of our schee and the output of reference decoder is negligible and better than the conventional ethod, and oreover PSNR gain by up to 8.66 db. Keywords : Thubnail, frequency doain, H.264/AVC core-transfor, rounding-off. a) VIA- VIA-Multiedia Center b) KT Central R&D Lab., KT, Korea : (shhong@edia.kw.ac.kr), 2010 (2010913),(20101019),(2010114) (HDTV: High Definition Television), (DMB: Digital Multiedia Broadcasting), TV(IPTV: Internet Protocol Television).
,, [1-2]. MPEG(Moving Picture Experts Group). MPEG.. [3-4].. DC DC, (thubnail).. MPEG [5-8]. H.264/AVC [9-10].. H.264/AVC [11-14]. H.264/AVC, [13]., H.264/AVC (Transfor and Quantization),. H.264/AVC. H.264/AVC (H.264/AVC Core Transfor), H.264/AVC... 2 H.264/AVC. 3 2. 4, 5.. H.264/AVC H.264/AVC.. 8 1 (a) 1. 4 4 (a) (b) 8 Fig. 1. Intra 4 4 prediction odes (a) Prediction block and its reference pixels (b) 8 direction of prediction 3 7 0 (b) 5 4 6
1. H.264/AVC H.264/AVC 16 16, 8 8, 4 4. 16 16 4 8 8, 4 4 9. H.264/AVC (High Profile) 8 8 (Baseline Profile) 16 16, 4 4. 1(a) ypred x1, x2, x3, x4. 1(b) 8 2(DC).. 16 16 4 4 [9]. 2. H.264/AVC H.264/AVC. X Q. (eleent ultiplication), dead zone,, H MF(ultiplication factor) H.264/AVC (core kernel). W x. H.264/AVC CT(H.264/ AVC Core Transfor). CTx W H xh T X Q. x. x H T X Q V H H T W H H V(scaling factor) H.264/AVC - (inverse core kernel). H.264/AVC [9-10]. 3. X Q HT Q x H xh T MF W MF H.264/AVC [13]. 2. Fig. 2. Conventional thubnail extractor
2. 2 H.264/AVC DC (DC Extraction). DC (4)(6) DC DC. Y pred HTx c X C Y pred Y pred x c X C n X n Q n n x c p x T X C P X T T C X X n C n C T X x n 1(a) y pred n x n, x n,. c n i. c ni x n x n HT. HT(H.264/AVC Transfor - 4 4 DCT) (7). HTx X H xh T PF. DC Y pred Y pred. 4. (4)(6) HT H.264/AVC. H.264/AVC HT (1) HT Q (H.264/AVC ). HT Q DC. H.264/AVC DCT X. MF PF H.264/AVC DCT PF MF. MF. H PF H.264/AVC (core kernel) - (post scaling factor) [9-10]. C ni X n X n c ni x n x n HT Q, C, P PF MF MF [16]. H.264/AVC X Q H xh T MF H xh T PF H xh T PF X
H.264/AVC. X. (1) (9). x. Y e. Y e HT. (4)(6). HT (11) (13) pred, pred., pred H.264/AVC... HT Q. y dc Y pred pred r Y e HT Q (0,0) (DC), 4. y dc 1/4. Y pred r, pred. Y e,. (14) H.264/AVC. H.264/AVC... Y pr ed X C HTy pr ed X C X C P X T pr ed Y pr ed Y pr ed n X n Q n T C X pr ed HTy pr ed n X n C n C T X pr ed HTy pr ed
III. 1. H.264/AVC H.264/AVC HT Q (1) (2) CT MF.. (4)(6). H.264/AVC [10].. CTxc H xch T H xh T OH ch T CTxOCTc.. H.264/AVC., (3) V W H.264/ AVC. x [15]. (3) H.264/AVC x. x T H X Q V H O diag H T OH I T H W H. (15) (4)(6). Y pred Y pred CTx c X OC P OX T CTy pred x c X OC x c X OC n X n OQ n T C OX p x T H PW PH T P diag H T P H I. (19) (2). CT x H xh T PW P W P Y pred CTy pred n X n OC n C T OX (12)(14) (16)(18) P P. P
3. Fig. 3. Proposed thubnail extractor of core transfor base H.264/AVC X Q H.264/AVC W (20) P x. 3. 2. H.264/AVC. a b c d y b c d e c d e f d e f g a y pred A B C. A, B, C 1 4 4. (22)..... 4 (0, 1, 2, 3) 1/4. [17]. y r (23) y r y pr ed.... y r. y r 2. 2. 2. y pred A B C D IJ K L AD, IL. (23).
y r (25) 2. y r u x n x n.. LUT. Y r CTy r Y r Vy r 여기에서y r Y r CTy r Y r Hy r 여기에서y r (27), (28) Y r,y r LUT. Y V, Y H. (27), (28) Y V, Y H. DC. 3. H.264/AVC 4. DC Fig. 4. The structure of the proposed DC extractor
. 4 DC.. Step 1) Y e O Y e Y e. Step 2) O LUT x n x n Y pred, Y pred, Y pred Y r, Y r, Y r. Step 3) Y pred Y r Y e DC 1/16. Step 4) Y V, Y H... REF(Reference) 4x4 16, 1/4. H.264/AVC. SDM(Spatial Doain Method) H.264/AVC, 1/4. HTM(H.264/AVC Transfor Method) HT(H.264/AVC ). CTM(Core Transfor Method) CT(H.264/ AVC ), CTMEC(Core Transfor Method with Error Copensation) CTM.. 1. 1. Table 1. Experiental Environent JM 12.2[18] BP, Level 4.0 / 22, 27, 32, 37 704 576 1920 1080 City, Crew, Harbour, Soccer Crowd run, Rush hour, Sunflower, Toys and Calendar 176 144 / 480 270 CTM 5. 5 SSE(Su of Squared Error). 5 2 HTM H.264/AVC. HTM /. 37 HTM
5. Fig. 5. The predicted residual block error after inverse quantization 0. AC. 2 (10) HTM AC. City(704 576) Crowdrun(1920 1080). CTM. CTM. 2 704 576 1920 1080 PSNR. PSNR REF HTM, CTM, 2. PSNR Table 2. Perforance coparison in PSNR for thubnail extraction ethods at each QP
CTMEC. 2 HTM CTM CTMEC 4.34 db 8.66 db PSNR. 2 PSNR.. 6 Crew 100, Crowd run 60 4 4. 6 16 16,. 2 PSNR. PSNR CTMEC.. 7 HTM CTM. 7 (a) HTM H.264/AVC. 7 (b) 6. Fig. 6. Intra prediction ode selection ratio per QP
(a) (b) 7. Rush hour (480 270, QP=37) (a) HTM, (b) CTM Fig. 7. Subjective quality perforance of Rush hour sequence (480 270, QP=37) (a) HTM, (b) CTM CTM. 811. 8 9 724 576 176 144. 10 11 1920 1080 480 270. 811 (b) HTM. 7 (a) (a) (b) (c) (d) 8. City (176 144, QP=22) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC Fig. 8. Subjective quality perforance of City sequence (176 144, QP=22) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC thubnail (a) (b) (c) (d) 9. Soccer (176 144, QP=27) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC Fig. 9. Subjective quality perforance of Soccer sequence (176 144, QP=27) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC thubnail
(a) (b) (c) 10. Sunflower (480 270, QP=37) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC Fig. 10. Subjective quality perforance of Sunflower sequence (704x576, QP=37) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC thubnail (d) (a) (b) (c) 11. Toys and calendar (480 270, QP=37) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC Fig. 11. Subjective quality perforance of Toys and calendar sequence (704x576, QP=37) (a) SDM, (b) HTM, (c) CTM, (d) CTMEC thubnail (d) H.264/AVC. 811 CTM (c) CTMEC
(d) H.264/AVC - SDM (a).. H.264/AVC, H.264/AVC. H.264/AVC., H.264/AVC, H.264/AVC (H.264/AVC Core Transfor).. 8.66 db PSNR... [1] B. Yeo and B. Liu, "Rapid scene analysis on copressed video", IEEE Trans. Circuit and Syste for video Technology, vol. 5, no. 6, pp.533-540, 1995. [2] Wing-San Chau, Oscar C. Au, Tak-Song Chong, Tai-Wai Chan, Chi-Shun Cheung, "Efficient Scene Change Detection in MPEG Copressed Video Using Coposite Block Averaged Luinance Iage Sequence", IEEE ICICS, pp. 688-691, 2005. [3],,,, vol. 10, no. 3, pp. 413-421, 2005. [4] N. Merhav and V. Bhaskaran, "Fast algoriths for DCT-doain iage down-sapling and for inverse otion copensation", IEEE Trans. Circuit and Syste for video Technology, vol. 7, pp. 468-476, 1997. [5] Taskiran, C. Jau-Yuen Chen, Bouan, C.A. Delp, E.J. "A copressed video database structured for active browsing and search," IEEE iage Processing, vol.3, pp.133-137, Oct. 1998. [6] Ja-Cheon Yoon, Hyeokan Ki, Seong Soo Chun, Jung-Ri Ki, Sanghoon Sull, "Real-Tie Video Indexing Syste for Live Digital Broadcast TV Progras," Lecture Notes in Coputer Science, CVIR2004, vol. 3115, pp. 261-269, Jul. 2004. [7] G.C. Feng and J. Jiang, "Iage extraction in DCT doain," IEE Proc.-Vis. Iage Signal Process., Vol. 150, No. 1, pp.20-27, Feb. 2003. [8] J. Song and B.L. Yeo, "Fast Extraction of Spatially Reduced Iage Sequences fro MPEG-2 Copressed Video," IEEE Trans. Circuits and Systes for Video Technology, vol. 9, no.7, pp. 1100-1114. Oct. 1999. [9] Draft ITU-T Recoendation and Final Draft International Standard of Joint Video Specification (ITU-T Rec. H.264 ISO/IEC 14496-10 AVC), 2003. [10] Henrique S. Malvar, Antti Hallapuro, Marta Karczewicz, and Louis Kerofsky, "Low-Coplexity Transfor and Quantization in H.264/AVC", IEEE Trans. Circuits and Systes for Video Technology, vol. 13, no. 7, pp. 598-603, July 2003. [11] Chen Chen, Ping-Hao Wu and Chen H, "Transfor-Doain Intra Prediction for H.264," IEEE ISCAS, pp.1497-1500, May 2005. [12],,,,,, H.264/AVC,, 13, 2, pp,222-235, Mar, 2008. [13] Eun-Seok Ki, Tae-Woong U, and Seoung-Jun Oh, "A Fast Thubnail Extraction Method in H.264/AVC Video Streas," IEEE Trans. Consuer Electronics, vol. 55, no. 3, pp. 1424-1430, Aug. 2009. [14],,,,, H.264/AVC,, pp,277-280, Nov, 2009 [15] Yu-Kuang Tu, Jar-Ferr Yang, and Ming-Ting Sun, "Efficient Rate-Distortion Estiation for H.264/AVC Coders", IEEE Trans. Circuits and Systes for Video Technology, vol. 16, no. 5, pp. 600-611, May 2006. [16] Jeong Mee Moon, and Jae Ho Ki, "A New Low-Coplexity Integer Distortion Estiation Method for H.264/AVC Encoder", IEEE Trans. Circuits and Systes for Video Technology, vol. 20, no. 2, pp. 207-212, Feb. 2010. [17] Yuichiro Nakaya, Yoshinori Suzuki, and Satoshi Misaka, "Avoidance of error accuulation in otion copensation," Doc. Q15-A-35R1, June 1997 [18] H.264/AVC Reference Software Model (JM 12.2) : http://iphoe.hhi. de/suehring/tl/download/old j
- 2009 2 : - 2009 3 ~ : - : H.264/AVC,, Thubnail - 2009 2 : - 2009 3 ~ : - : H.264/AVC, - 1983 : - 1985 : - 1989 : - 1989 ~ : KT - : IPTV, /STB, CAS/DRM,, UI/UX - 1986 : - 1988 : KAIST - 2003 : KAIST - 1988 ~ : KT - :,,, - 1980 2 : () - 1982 2 : () - 1988 5 : Syracuse University / () - 1982 3 ~ 1992 8 : ( ) - 1986 7 ~ 1986 8 : NSF Supercoputer Center - 1987 5 ~ 1988 5 : Northeast Parallel Architecture Center - 1992 3 ~ 1992 8 : - 1992 9 ~ : (VIA-Multiedia Center ) - 2002 3 ~ : SC29-Korea MPEG Foru - :,,,