(Special Paper) 22 5, 2017 9 (JBE Vol. 22, No. 5, Sepember 2017) https://doi.org/10.5909/jbe.2017.22.5.541 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) (JEM) a), a), a) A Fast Decision Method of Quadtree plus Binary Tree () Depth in JEM Yong-Uk Yoon a), Do-Hyun Park a), and Jae-Gon Kim a) JVET(Joint Video Exploration Team) SW JEM(Joint Exploration Model) (Quadtree plus Binary Tree) (CU).,. JEM (depth) -(Rate-Distortion: RD). JEM 5.0 AI(All ) 0.7% BD-rate 21.6%, RA(Random Access) 1.2% BD-rate 11.0%. Abstract The Joint Exploration Model (JEM), which is a reference SW codec of the Joint Video Exploration Team (JVET) exploring the future video standard technology, provides a recursive Quadtree plus Binary Tree () block structure. can achieve enhanced coding efficiency by adding new block structures at the expense of largely increased computational complexity. In this paper, we propose a fast decision algorithm of block partitioning depth that uses the rate-distortion (RD) cost of the upper and current depth to reduce the complexity of the JEM encoder. Experimental results showed that the computational complexity of JEM 5.0 can be reduced up to 21.6% and 11.0% with BD-rate increase of 0.7% and 1.2% in AI (All ) and RA (Random Access), respectively. Keyword : future video coding, JVET, JEM,, fast depth decision, fast encoding a) (Korea Aerospace University, School of Electronics and Information Engineering) Corresponding Author : (Jae-Gon Kim) E-mail: jgkim@kau.ac.kr Tel: +82-2-300-0414 ORCID: http://orcid.org/0000-0003-3686-4786 2017 () (No. 2016-0-00572, HEVC/3DA 2 5 / ) 29 (IPIU 2017). Manuscript received April 17, 2017; Revised May 17, 2017; Accepted May 17, 2017. 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 (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. 22, No. 5, Sepember 2017). H.264/AVC 2 HEVC, UHD. 5G 360 VR(Virtual Reality). ITU-T VCEG ISO/IEC MPEG JVET(Joint Video Exploration Team) 2020 HEVC 2. JVET JEM (Joint Exploration Model) 5.0. JEM 5.0 HEVC HEVC 27% 10 [1][2]. HEVC, CTU(Coding Tree Unit), CTB(Coding Tree Blocks) CTB [3]. CTU (CU: Coding Unit) (PU: Prediction Unit), (Transform Unit). JEM HEVC CU, PU, TU 1 CU. JEM CTC(Common Test Condition) 64 64 CTU 128 128 CTU (Quad Tree) 4, (Binary Tree) 3 (Depth) [4]. CU., -, CU. HEVC H.264/AVC, [4~6]. JVET, CU SKIP DEPTH, -, [7~10]. CU - CU - CU. 1. JEM Fig. 1. block structure in JEM. 1. JEM CU 2. CU - CU. CU. 2. Fig. 2. Recursive block partition -. -
3. - Fig. 3. Comparison RD-cost of upper depth with current depth., - -. 3 - - 1/4 -, - - 1/4. 2. ( ) CU (Ref), - -., CU - 1/4( 1/2) - -. 4. - - 1 6. - 1/4 - (1.2 1/4Upper < Cur), - 4. Fig. 4. Method of depth application according to reference block 1. - 6 Case Table 1. 6-Case according to RD-cost Condition Case Case 1 Case 2 Case 3 Case 4 Case 5 Case 6
(JBE Vol. 22, No. 5, Sepember 2017) 5. - Case (, ) Fig. 5. Probability of occurrence according to each case of RD cost. 5 1 Case, JEM A, B 4.,. 2 3 1 Case, 5,,., Case 2 3%., 5 Case 2 (37%) 2 Case 2 (9%). - Case 2 3%. 3% Case 2., Case 1 6%. 2. Case (, /) Table 2. Depth variation for each case (, /) Rate(%) Prob(%) Rate(%) Prob(%) Rate(%) Prob(%) Rate(%) Prob(%) 5 1 3 0 24 9 16 2 Case1 58 10 76 3 Case2 67 25 76 8 37 6 21 1 9 3 8 1 17 1 47 5 4 0 3 0 Case3 79 5 46 4 Case4 63 3 42 4 5 0 7 1 33 2 56 5 5 0 7 1 7 2 18 9 Case5 75 5 66 9 Case6 88 24 71 36 20 1 26 4 5 1 11 6 3. Case (Inter, /) Table 3. Depth variation for each case (Inter, /) Rate(%) Prob(%) Rate(%) Prob(%) Rate(%) Prob(%) Rate(%) Prob(%) 5 0 0 0 14 9 0 0 Case1 41 0 0 0 Case2 71 25 0 0 55 0 0 0 14 3 0 0 31 4 36 5 3 0 5 1 Case3 66 9 60 9 Case4 62 9 54 8 3 0 4 1 35 5 41 6 4 1 9 1 17 8 16 9 Case5 64 15 67 9 Case6 76 35 74 43 33 8 24 3 7 3 9 5
. (1)., (1) - 1 2.. JEM 5.0. JEM A1, A2, B QP 22, 27, 32, 37 [11]. 4 Main Profile AI, RA 5. JEM 5.0 Anchor (2). 4 AI JEM 5.0 0.7% BD-rate 21.6%, RA 1.2% BD-rate 11.0%. B class A class. CTU, CTU.. 4. - (AI, Anchor: JEM 5.0) Table 4. Experimental results of fast depth decision using RD-cost (AI, Anchor: JEM 5.0) Class Sequence Y U V A2 Tango 0.5% -0.1% 0.1% -29% 4096 2160 A1 3840 2160 B 1920 1080 Rollercoaster 0.5% 0.1% 0.1% -26% Average 0.5% 0.0% 0.1% -27.5% CatRobot 0.6% -0.4% -0.1% -14% Drums 0.5% 0.3% 0.2% -31% Average 0.6% -0.1% 0.1% -22.5% BasketballDrive 1.3% 0.2% 0.2% -16% Cactus 0.9% 0.1% 0.1% -14% Average 1.1% 0.2% 0.2% -15% Total Average 0.7% 0.0% 0.1% -21.6% 5. - (RA, Anchor: JEM 5.0) Table 5. Experimental results of fast depth decision using RD-cost (RA, Anchor: JEM 5.0) Class Sequence Y U V A2 Tango 1.2% 1.5% 2.0% -13% 4096 2160 A1 3840 2160 B 1920 1080 Rollercoaster 1.2% 1.9% 1.9% -13% Average 1.2% 1.7% 2.0% -13% CatRobot 1.3% 1.2% 1.5% -11% Drums 1.5% 1.4% 1.2% -14% Average 1.4% 1.3% 1.4% -12.5% BasketballDrive 0.7% -0.2% 1.3% -7% Cactus 1.2% 0.6% 1.7% -8% Average 1.0% 0.2% 1.5% -7.5% Total Average 1.2% 1.1% 1.6% -11%
(JBE Vol. 22, No. 5, Sepember 2017) 6. Fig. 6. An example result of fast block partition method 6 JEM 5.0. CU... JEM -. 3 -.. JEM 5.0 AI 0.7% BD-rate 21.6%, RA 1.2% BD-rate 11.0%.,. (References) [1] J. Chen, E. Alshina, G. J. Sullivan, J. -R. Ohm, and J. Boyce, Algorithm Description of Joint Exploration Test Model 5, JVET document, JVET-E1001, Jan. 2017. [2] X. Li, K.Suehring, Report of AHG3 on JEM software development, JVET document, JVET-E0003, Jan. 2017. [3] I. Kim, J. Min, T. Lee, W. Han, Block Partitioning Structure in the HEVC Standard, IEEE Trans. Circuits. Syst. Video Technol., vol. 22, no. 12, pp. 1697-1706, Dec. 2012. [4] J. G. Jang, H. Y. Choi, and J. G. Kim, Fast PU Decision Method Using Coding Information of Co-Located Sub-CU in Upper Depth for HEVC, J. Broadcasting Engineering, vol. 20, no. 2, Mar. 2015. [5] J. W. Kim, D. H. Kim, J. G. Kim, An Early Termination Algorithm of Prediction Unit (PU) Search for Fast HEVC Encoding, J. Broadcasting Engineering, vol. 19, no. 5, Sep. 2014. [6] H. M. Yoo, S. Y. Shin, J.W. Suh, Fast CU Decision Algorithm using the Initial CU Size Estimation and PU modes RD Cost, J. Broadcasting Engineering, vol. 19, no. 3, May. 2014. [7] J. An, H. Huang, K. Zhang, Y. W. Huang, S. Lei, Quadtree plus binary tree structure integration with JEM tools, JVET document, JVET-B0023, Feb. 2017. [8] Y. Yamamoto, T. Ikai, Y. Yasugi, AHG5: Improved fast encoding setting, JVET document, JVET-E0023, Jan. 2017. [9] P. -H. Lin, Y. -J. Chang, C. -L. Lin, C. -C. Lin, AHG5: Improved fast algorithm in JEM 4.0, JVET document, JVET-E0078, Jan. 2017. [10] Z. Wang, S. Wang, J. Zhang, S. Ma, Local-constrained quadtree plus binary tree block partition structure for enhanced video coding, In. Proc. Visual Communications and Image Processing (VCIP), 2016. [11] K. Suehring, X. Li, JVET common test conditions and software reference configurations, JVET document, JVET-B1010, Feb. 2017.
- 2017 2 : - 2017 3 ~ : - ORCID : http://orcid.org/0000-0002-5105-5437 - :,, 360 VR - 2016 2 : - 2016 3 ~ : - ORCID : http://orcid.org/0000-0002-5873-0132 - :,, 360 VR - 1990 2 : - 1992 2 : KAIST - 1992 2 : KAIST - 1992 3 ~ 2007 2 : (ETRI) / - 2001 9 ~ 2002 7 : - 2015 12 ~ 2016 1 : UC San Diego, Visiting Scholar - 2007 9 ~ : - ORCID : http://orcid.org/0000-0003-3686-4786 - :,,,