AVS2视频编码标准技术特色及应用

doi:10.11959/j.issn.1000-0801.2017245

摘要/Abstract

摘要：

围绕视频编码核心技术，简单介绍AVS视频编码标准发展历程，详细介绍了最新一代AVS视频编码标准——AVS2（GB/T 33475.2-2016）的核心创新技术，主要包括：灵活的预测划分方式、多假设帧间预测、优化的层次变换设计和自适应环路滤波处理技术以及面向场景视频应用 AVS2 提出的基于场景建模的高效预测编码技术。和第一代AVS视频编码标准相比，AVS2编码效率提升一倍以上；和同期HEVC/H.265国际标准相比， AVS2在场景视频编码方面有显著优势。另外简单介绍了AVS标准在数字电视广播等行业中的应用情况。

关键词: AVS, 变换, 预测, 熵编码, 环路滤波

Abstract:

A brief introduction was given to the development history of AVS video coding standards.Then an overview of the key technologies adopted in AVS2 video coding standard (GB/T 33475.2-2016) was provided,including flexible prediction partition,multiple hypothesis prediction,optimized two-level transform,adaptive loop filter and background picture model based prediction coding for scene video.Compared to the AVS1,AVS2 can achieve more than 50% bits saving.And compared to HEVC/H.265,AVS2 can achieve significant coding efficiency improvement for scene video coding.Moreover,the applications of AVS standard in digital TV broadcasting was introduced briefly.

Key words: AVS, transform, prediction, entropy coding, loop filter

中图分类号:

TN919.81

马思伟,罗法蕾,黄铁军. AVS2视频编码标准技术特色及应用[J]. 电信科学, 2017, 33(8): 1-15.

Siwei MA,Falei LUO,Tiejun HUANG. Kernel technologies and applications of AVS2 video coding standard[J]. Telecommunications Science, 2017, 33(8): 1-15.

图/表 22

图1

表1

图2

图3

图4

图5

图6

图7

图8

图9

图10

图11

图12

图13

表2

表3

表4

表5

表6

表7

图14

图15

参考文献 27

[1]	朱秀昌 . 视频编码新标准——H.264[J]. 电信科学, 2002,18(12): 26-29.
	ZHU X C . A new video coding recommendation——H.264[J]. Telecommunications Science, 2002,18(12): 26-29.
[2]	解伟, 郭晓强, 全子一 . IPTV 中的视频压缩技术研究[J]. 电信科学, 2006,22(3): 39-42.
	XIE W , GUO X Q , QUAN Z Y . Research on video compression in IPTV application[J]. Telecommunications Science, 2006,22(3): 39-42.
[3]	郭春辉, 周经野, 刘华东 ,等. AVS1-P7与H.264关键技术及性能比较[J]. 电信科学, 2006,22(1): 47-50.
	GUO C H , ZHOU J Y , LIU H D ,et al. Key technologies and performance comparison of AVS1-P7 with H.264[J]. Telecommunications Science, 2006,22(1): 47-50.
[4]	CUTLER C . Differential quantization of communication signals:US patent 2605631[S]. 1952.
[5]	OLIVER B . Efficient coding[J]. The Bell System Technical Journal, 1952(31): 724-750.
[6]	HARRISON C . Experiments with linear prediction in television[J]. The Bell System Technical Journal, 1952,31(7): 764-783.
[7]	CANDY J , FRANKE M , HASKELL B ,et al. Transmitting television as clusters of frame-to-frame differences[J]. The Bell System Technical Journal, 1971,50(7): 1889-1917.
[8]	NETRAVALI A , ROBBINS J . Motion-compensated television coding:part 1[J]. The Bell System Technical Journal, 1979,58(3): 631-670.
[9]	AHMED N , NATARAJAN T , RAO K R . Discrete cosine transform[J]. IEEE Transactions on Computers, 1974,C(32): 90-93.
[10]	MALVAR H , HALLAPURO A , KARCZEWICZ M ,et al. Low-complexity transform and quantization in H.264/AVC[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2003,13(7): 598-603.
[11]	NARROSCHKE M . Coding efficiency of the DCT and DST in hybrid video coding[J]. IEEE Journal of Selected Topics in Signal Processing, 2013,7(6): 1062-1071.
[12]	TESCHER A G , COX R V . An adaptive transform coding algorithm[C]// 1976 IEEE International Conference on Communications,June 14-16,1976,Philadelphia,USA. New Jersey:IEEE Press, 1976.
[13]	KOGA T , IINUMA K , HIRANO A ,et al. Motion-compensated inter-frame coding for video conferencing:NTCG5.3.1-G 5.3.5[R]. 1981.
[14]	马思伟 . AVS 视频编码标准技术回顾及最新进展[J]. 计算机研究与发展, 2015,52(1): 27-37.
	MA S W . History and recent developments of AVS video coding standards[J]. Journal of Computer Research and Development, 2015,52(1): 27-37.
[15]	余全合, 曹潇然, 李蔚然 ,等. 短距离帧内预测技术:AVS_M3171[S]. 2013.
	YU Q H , CAO X R , LI W R ,et al. Short range intra prediction technique:AVS_M3171[S]. 2013.
[16]	FLIERL M , WIEGAND T , GIROD B . Rate-constrained multihypothesis prediction for motion compensated video compression[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2002,12(11): 957-969.
[17]	凌勇, 虞露 . F帧CE:一种前向双假设预测图像块编码技术:AVS_M3326[S]. 2014.
	LING Y , YU L . F frame CE:a block image coding technique based on forward two hypothesis prediction:AVS_M3326[S]. 2014.
[18]	KIM I K , LEE S , PIAO Y J ,et al. Directional multi-hypothesis prediction (DMH) for AVS2:AVS_M3094[S]. 2013.
[19]	李蔚然, 袁媛, 曹潇然 ,等. 非平方的四叉树变换:AVS_M3153[S]. 2013.
	LI W R , YUAN Y , CAO X R ,et al. Non-square quad-tree transform:AVS_M3153[S]. 2013.
[20]	林和源, 童怡新, 王颂文 . 帧内编码二次转换改进提案:AVS_M3296[S]. 2014.
	LIN H Y , TONG Y X , WANG S W . Improvement proposal of the second time transform of intraframe coding:AVS_M3296[S]. 2014.
[21]	RISSANEN J . Generalized draft inequality and arithmetic coding[J]. IBM Journal of Research and Development, 1976(20): 198-203.
[22]	MARPE D , SCHWARZ H , WIEGAND T.Context-based adaptive binary arithmetic coding in the H . 264/AVC video compression standard[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2003,13(7): 620-636.
[23]	WANG J , WANG X , JI T ,et al. Two-level transform coefficient coding:AVS_M3035[S]. 2012.
[24]	LEE J S , KIM C , FU C ,et al. Sample adaptive offset for AVS2:AVS_M3197[S]. 2013.
[25]	ZHANG X , SI J , WANG S ,et al. Adaptive loop filter for AVS2:AVS_M3292[S]. 2014.
[26]	ZHANG X , HUANG T , TIAN Y ,et al. Low-complexity and high-efficiency background modeling for surveillance video coding[C]// 2012 Visual Communications and Image Processing (VCIP 2012),Nov 27-30,2012,San Diego,USA. New Jersey:IEEE Press, 2012.
[27]	BOSSEN F . Common HM test conditions and software reference con-figurations:ITU-T SG16 ContributionJCTVC-L1100[S]. 2013.

技术完成时间	AVS／类	应用及主要技术特征
2003年12月	AVS1／基准	面向数字电视广播应用，采用基于8×8像素块的帧内预测，8×8像素块变换编码，去块效应滤波，变块大小运动补偿（8×8像素～16×16像素）
2008年6月	AVS1／伸展	面向监控视频应用，采用基于背景帧、核心帧的编码技术
2008年9月	AVS1／加强	面向数字电影应用，采用基于上下文的算术编码、加权量化等技术
2009年6月	AVS1移动	面向移动视频应用，采用了4×4像素帧内预测、8×8像素/4×4像素自适应变换等技术
2011年7月	AVS1监控	增强的监控视频编码，采用背景图像建模编码技术
2012年5月	AVS+／广播	面向高清数字电视广播，采用了基于上下文的算术编码、帧级加权量化、增强场预测编码等技术
2015年12月	AVS2／基准	面向超高清数字电视广播、场景视频等应用，采用了自适应预测划分、多假设预测、层次变换、自适应算术编码、自适应滤波等技术

配置序列	RA			LD			AI
配置序列	Y	U	V	Y	U	V	Y	U	V
UHD	-50.5%	-56.3%	-56.2%	-57.6%	-66.5%	-67.1%	-31.2%	-30.8%	-30.5%
1080P	-51.3%	-61.7%	-63.2%	-44.3%	-59.0%	-62.1%	-33.1%	-37.4%	-39.2%
WVGA	-52.8%	-58.1%	-58.6%	-50.5%	-58.1%	-59.9%	-30.4%	-28.2%	-29.8%
WQVGA	-52.4%	-58.9%	-59.0%	-49.4%	-59.2%	-58.7%	-26.6%	-24.8%	-26.9%
720P	-57.2%	-66.1%	-63.8%	-56.3%	-71.1%	-69.3%	-34.0%	-37.7%	-34.8%
平均	-52.9%	-60.5%	-60.5%	-51.0%	-62.1%	-62.8%	-31.2%	-32.4%	-33.0%
编码复杂度		1 210%			2 102%			1 228%

配置序列	RA			LD			AI
配置序列	Y	U	V	Y	U	V	Y	U	V
UHD	-0.3%	5.6%	6.3%	2.7%	4.8%	9.0%	-2.2%	2.1%	2.1%
1080P	-2.3%	5.6%	3.8%	0.7%	-2.4%	-1.6%	-0.7%	2.1%	1.2%
WVGA	0.0%	6.7%	7.7%	0.9%	2.2%	5.2%	1.5%	3.9%	3.8%
WQVGA	1.1%	7.7%	8.8%	4.9%	8.1%	9.8%	2.8%	4.5%	4.9%
720P	-2.4%	-7.3%	-5.5%	1.9%	-5.1%	0.3%	-2.1%	-5.3%	-4.2%
平均	-0.9%	4.1%	4.6%	2.1%	1.3%	4.1%	-0.1%	1.4%	1.4%
编码复杂度		315%			312%			327%

配置序列	RA			LD
配置序列	Y	U	V	Y	U	V
Crossroad_720×576_30	-39.2%	-37.3%	-37.7%	-25.6%	-64.6%	-60.4%
Office_720×576_30	-24.0%	-24.7%	-25.0%	-15.5%	-46.1%	-44.3%
Overbridge_720×576_30	-60.7%	-56.9%	-56.8%	-38.4%	-58.3%	-56.2%
Intersection_1 600×1 200_30	-18.7%	-29.0%	-28.4%	-23.2%	-33.7%	-29.6%
Mainroad_1 600×1 200_30	-52.7%	-65.2%	-58.6%	-53.8%	-58.9%	-54.3%
平均	-39.1%	-42.6%	-41.3%	-31.3%	-52.3%	-49.0%
编码复杂度		334%			332%

项目序列	配置	超快速（配置0）	比较快速（配置3）	中速（配置5）	慢速（配置7）	平和（配置9）
BasketballDrive 1 920 dpi×1 080 dpi,50帧/s	Y BD-Rate	11.36%	-12.55%	-18.98%	-4.29%	-11.15%
	加速比(x265)	14.02	7.02	4.10	0.64	0.10
	加速比(CAVS2)	8.75	3.80	1.32	0.75	0.12
BQTerrace 1 920 dpi×1 080 dpi,60帧/s	Y BD-Rate	-26.95%	-4.19%	-10.93%	-3.67%	-3.59%
	加速比(x265)	19.46	11.65	5.68	0.96	0.12
	加速比(CAVS 2)	10.00	5.83	2.03	0.99	0.11
Cactus 1 920 dpi×1 080 dpi,50帧/s	Y BD-Rate	-7.25%	-15.63%	-19.48%	-9.06%	-14.10%
	加速比(x265)	16.23	8.06	4.95	0.75	0.11
	加速比(CAVS2)	10.28	4.37	1.56	0.83	0.12
Kimino 1 920 dpi×1 080 dpi,24帧/s	Y BD-Rate	11.48%	-6.57%	-14.27%	-6.63%	-9.06%
	加速比(x265)	13.78	6.41	3.91	0.79	0.12
	加速比(CAVS2)	10.44	4.94	1.77	0.94	0.14
ParkScene 1 920 dpi×1 080 dpi,24帧/s	Y BD-Rate	-6.87%	-8.63%	-11.55%	-5.33%	-6.25%
	加速比(x265)	16.09	7.91	4.71	0.84	0.12
	加速比(CAVS2)	9.54	4.65	1.73	0.92	0.12
平均	Y BD-Rate	-3.65%	-9.51%	-15.04%	-5.80%	-8.83%
	加速比(x265)	15.92	8.21	4.67	0.80	0.11
	加速比(CAVS2)	9.80	4.72	1.68	0.89	0.12