(Regular Paper) 23 1, 2018 1 (JBE Vol. 23, No. 1, January 2018) https://doi.org/10.5909/jbe.2018.23.1.115 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) HEVC a), a) Fast Partition Decision Using Rotation Forest for Intra-Frame Coding in HEVC Screen Content Coding Extension Jeonghwan Heo a) and Jechang Jeong a) High Efficiency Video Coding (HEVC) Screen Content Coding (SCC). HEVC... SCC 3.11% BD-BR 42%. Abstract This paper presents a fast partition decision framework for High Efficiency Video Coding (HEVC) Screen Content Coding (SCC) based on machine learning. Currently, the HEVC performs quad-tree block partitioning process to achieve optimal coding efficiency. Since this process requires a high computational complexity of the encoding device, the fast encoding process has been studied as determining the block structure early. However, in the case of the screen content video coding, it is difficult to apply the conventional early partition decision method because it shows different partition characteristics from natural content. The proposed method solves the problem by classifying the screen content blocks after partition decision, and it shows an increase of 3.11% BD-BR and 42% time reduction compared to the SCC common test condition. Keyword : HEVC, Screen Content Coding, Rotation Forest, Decision Tree Learning a) (Department of Electronics and Computer Engineering, Hanyang University) Corresponding Author : (Jechang Jeong) E-mail: jjeong@hanyang.ac.kr Tel: +82-2-2220-4370 ORCID: http://orcid.org/0000-0002-3759-3116. Manuscript received November 10, 2017; Revised November 30,2017; Accepted December 5, 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. 23, No. 1, January 2018). ( ). High Efficiency Video Coding (HEVC). 2014 1 ISO/IEC Moving Picture Expert Group, ITU-T Video Coding Expert Group Joint Collaborate Team on Video Coding (JCT-VC) Call for Proposal (CfP) [1]. JCT-VC Screen Content Test Model software Ver. 8.5 (SCM 8.5 [2] ) HEVC 55% BD-BR. (Intra Block Copy: IBC), (Palette Coding Mode: PLT), (Adaptive Color Transform: ACT), (Adaptive Motion Compensation Precision: AMCP) [3].. SCM Coding Unit (CU). SCM HEVC.. HEVC HEVC Screen Content Coding (SCC). SCC CU [4]. 4. HEVC. A. CU : CU HEVC. [5], [6], Coded Block Flag(CBF) [7]. Support Vector Machine (SVM) [8], (Neural Network: NN) [9], [10] CU. B. : [11]. [12]. C. : Hadamard Cost [13]. D. : (Motion Estimation). [14] Partial Distortion Search (PDS). HEVC.,,,,.,
,. SCC [15], NN [4]. (Rotation Forest). CU... 2.. 3. 4. 5. 1. HEVC ( 3 ) Fig. 1. Quadtree partition coding order of HEVC (three step of maximum depth) CU. CU. 2. (Inter Prediction). 1. (a) 1 HEVC. 64 64 Coding Tree Unit (CTU) 4 CU CU.. 4 CU CU Cost. (b) 2. (a) (b) Fig. 2. (a) Reference area for global block vector search (b) Reference area for local block vector search
(JBE Vol. 23, No. 1, January 2018). 2 (Block Vector).,... 2 (b) CTU. 1. _ 2 (a) CU. 16 16 CU, 1 1. 8 8 CU. 8 8 8 8.. (DC) (H). -... 3. (Palette Coding mode),. (Palette) (Indices Map). 3 RGB YUV (Palette Index). CU (Palette Predictor)... (2), 4 4,,. 3. Fig. 3. Index map and Palette in Palette mode encoding 4.
(Regression) (Cla- ssification).. 4 (Node) (Terminal Node), 2. (Squared Sum of Error) (Cross- Entropy) 3 (Gini Impurity). 4. Fig. 4. A simple structure of decision tree (Class Label),. 2. 4..,. (Training Set). (Validation Set).,.. (Feature).. (Subset) Principal Compo- nent Analysis (PCA). [16].. 1 : (n N N n ) : (N 1 ) : ( ) : (.) : :. (, For, ) For.. 75%. PCA..
(JBE Vol. 23, No. 1, January 2018). (Feature Selection),. 1. 5 3. (Feature Extraction) 1, 2. SCM CU., CU. CU CU. CU 1 2.,. CU 1. 2. 1 CU (Screen Content Blocks: SCB) (Natural Content Blocks: NCB). SCB,, CU. NCB 2 CU. 2 CU (Split-Class) (Unsplit-Class). S-Class RDO CU. U-Class CU CU. 2 HEVC. 2. (Feature Selection) 5. Fig. 5. Proposed encoder flowchart
SCB NCB, S-Class U-Class. 6. CU 3. CU. A. CU Gradient: CU Gradient. CU Gradient CU (6).. B. CU Partition Impurity: CU CU... CU 8 8 Gradient Gradient (500). 6 (a) 8 8 (Complexity Map) 6 (b) CU 4 CU. 0 (64 64 4 32 32 ) CU CU CU. 6 CU Partition Impurity. 6. (a) CU (b) Fig. 6. (a) CU partition impurity map (b) Example of partition impurity C. Intra Mode Cost:., Angular Mode. 7 35. min D. CU Color Count: CU RGB YUV.. E. Zero Gradient Count: Gradient 0. Zero Gradient Count. F. Gradient Peak: CU Gradient.
(JBE Vol. 23, No. 1, January 2018) max. 8. 3. (Soft Decision) (85%). 7 2.. 7 0.13 S-Class 86%. S-Class. x,, x,. x x 1 CU. 2 CU. 4.. Logistic Regression, Linear Perceptron Classifier, NN, SVM, NN SVM, [4].. 5. 7. Fig. 7. Soft-decision of rotation forest classifier CU (Supervised learning). (Common Testing Condition: CTC) [17] (Sequence) 10 SCM. CU.. CU QP CU, QP..
0.5%.. All Intra CTC, HM-16.15 SCM-8.5 [2], Intel Xeon E5-2690 @ 3.00GHz. BjØntegaard-Delta [18], ATS (Average Time Saving). Rotation Forest R Recursive Partitioning and Regression Trees (Rpart) library (Ver. 4.1-11). 500 85%. 1. Table 1. Classification accuracy of Rotation Forest Classifier Classifier 1 Classifier 2 QP CU Depth 64 64 32 32 16 16 8 8 22 87.3 90.9 78.2 81.3 27 88.2 90.5 80.3 81.9 32 89.9 90.6 80.0 83.6 37 88.8 89.8 79.7 84.6 22 93.6 84.7 77.5 * 27 93.5 84.7 80.1 * 32 91.6 84.6 82.1 * 37 90.5 84.3 82.5 * *8 8 block is lowest depth CU that is always pruned 0.4% 1. QP. 1 8 8 2 16 16 QP. QP. 6 (Feature). 2 2. Table 2. Experimental results of the proposed fast partition decision method Type Sequences Resolution BD-BR BD-PSNR ATS(%) CC Kimono 1080p 1.051-0.021 79.9 M MissionControlClip3 1080p 2.884-0.286 41.5 TGM Console 1080p 3.466-1.031 34.3 TGM Desktop 1080p 3.651-1.197 41.9 TGM FlyingGraphics 1080p 1.800-0.271 17.2 TGM Map 720p 2.119-0.220 42.0 TGM Programming 720p 3.059-0.327 32.1 A Robot 720p 2.022-0.084 62.3 TGM SlideShow 720p 6.795-0.552 40.0 TGM Web_browsing 720p 2.553-0.463 25.0 M BasketballScreen 1440p 4.399-0.393 45.4 M MissionControl2 1440p 3.530-0.311 46.7 Average 3.111-0.430 42.4 TGM: Text and Graphics with Motion MC: Mixed Content A: Animation CC: Camera-Captured Content
(JBE Vol. 23, No. 1, January 2018). Kimono, Slide- Show. Kimono 8 8. 1 NCB. SlideShow SCB. SlideShow. CTC. Console, Desktop BD-BR.. 8 8. 3. BD-BR. 3. Table 3. Experimental results of the proposed fast partition decision method Previous Fast Encoder works Codebase BD-BR ATS(%) Tsang, Chan and Siu [12] SCM-2.0 0.66 29% Duanmu, Ma, Wang [4] SCM-4.0 3.69 36% Proposed algorithm SCM-8.5 3.11 42%.. HEVC CU.. CU,. 3.11% BD-BR 42%.,.. (References) [1] ISO/IEC JTC1/SC29/WG11 and ITU-T Q6/SG16, MPEG2014/ N14175/VCEGAW90, Joint Call for Proposals for Coding of Screen Content, San Jose, USA, Jan. 2014. [2] HM-16.16+SCM-8.5 Software, https://hevc.hhi.fraunhofer.de/trac/ hevc/browser/tags/hm-16.16%2bscm-8.5 (accessed Nov. 01, 2016). [3] Joshi, J. Xu, R. Cohen, S. Liu, Y. Ye (editors) High Efficiency Video Coding (HEVC) Range Extensions text specification: Draft 7, Document JCTVC-Q1005, in ITU-T SG16 WP3 and ISO/IEC JTC1/SC29/WG11, Apr. 2014. [4] F. Duanmu, Z. Ma, and Y. Wang, Fast CU partition decision using machine learning for screen content compression, in IEEE Int. Conf. Image Process. (ICIP), pp.49724976, Sep. 2015. [5] J. Jang, H. Choi, and J. Kim, Fast PU Decision Method Using Coding Information of Co-Located Sub-CU in Upper Depth for HEVC, Journal of Broadcast Engineering, Vol.20, No.2, pp.340-347, Mar 2015. [6] D. Lee, and J. Jeong. Fast intra coding unit decision for high efficiency video coding based on statistical information, Elsevior Signal Processing Image Communication Vol. 55, pp. 121-129, July. 2017. [7] S Jeon, N kim, and B Jeon, CU Depth Decision Based on FAST Corner Detection for HEVC Intra Prediction, Journal of Broadcast Engineering, Vol.21, No.4, pp.484-492, July 2016. [8] Y. Zhang, S. Kwong, L. Xu, and G. Jiang, DIRECT mode early decision optimization based on rate distortion cost property and interview correlation, IEEE Trans. Broadcast, vol. 59, no. 2, pp. 390398, Jun. 2013. [9] J. Chiang, W. Chen, L. Liu, K. Hsu, and W. Lie, A fast H.264/AVC based stereo video encoding algorithm based on hierarchical two-stage
neural classification, IEEE Signal Process, vol. 5, no. 2, pp. 309320, Apr. 2011. [10] S. Ryu and J. Kang, Machine-Learning based Fast Intra Mode Decision Algorithm in HEVC, International Technical Conf. on Circuits/Systems, Computers and Communication (ICT-CSCC), 2017. [11] W. Han, J. Ahn, J. Lee, Early Decision of Inter-prediction Modes in HEVC Encoder, Journal of Broadcast Engineering, Vol.20, No.1, pp.171-182, Jan 2015. [12] S. Tsang, Y. Chan, and W. Siu, Fast and efficient intra coding techniques for smooth regions in screen content coding based on boundary prediction samples, in Proc. ICASSP, pp. 14091413, 2015. [13] Y. Piao, J. Min, and J. Chen, Encoder Improvement of Unified Intra Prediction, document JCTVC-C207, Jan. 2013. [14] C. Bei and R. Gray, An improvement of the minimum distortion encoding algorithm for vector quantization, IEEE Trans. Commun, vol. COM-33, pp. 11321133, Oct. 1985. [15] D. K. Kwon and M. Budagavi, Fast intra block copy (IntraBC) search for HEVC screen content coding, in IEEE International Symposium on Circuits and Systems (ISCAS), Melbourne VIC, 2014, pp. 9-12. [16] J. Rodriguez, L. Kuncheva and C. Alonso, Rotation Forest: A New Classifier Ensemble Method, in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28, no. 10, pp. 1619-1630, Oct. 2006. [17] H. Yu, R. Cohen, K. Rapaka, and J. Xu, Common Test Conditions for Screen Content Coding, document JCTVC-T1015, Feb. 2015. [18] G. Bjontegaard, Calculation of average PSNR differences between RD curves, Video Coding Experts Group (VCEG), VCEG-M33, Austin, Texas, U.S.A., April, 2001. - 2015 8 : () - 2015 8 ~ : () - ORCID : https://orcid.org/0000-0001-8670-9800 - : (HEVC),, - 1980 2 : () - 1982 2 : KAIST () - 1990 : () - 1980 ~ 1986 : KBS ( ) - 1990 ~ 1991 : ( ) - 1995 ~ : ( ) - 1990 12 : - 1998 11 : - 2007 : IEEE Chester Sall Award - 2008 : ETRI Journal Paper Award - 2011 5 : 46 - ORCID : http://orcid.org/0000-0002-3759-3116 - :,, 3DTV