文献类型：期刊文献

中文题名：通用串预测算法及在AVS2屏幕与混合内容视频编码中的应用

英文题名：A Universal String Prediction Approach and Its Application in AVS2 Mixed Content Coding

作者：赵利平[1,2,3];周开伦[2];林涛[2];郭靖[2]

机构：[1]绍兴文理学院计算机科学与工程系,浙江绍兴312000;[2]同济大学超大规模集成电路研究所,上海200092;[3]嘉兴学院数理与信息工程学院,浙江嘉兴314000

年份：2019

卷号：42

期号：9

起止页码：2100

中文期刊名：计算机学报

外文期刊名：Chinese Journal of Computers

收录：CSTPCD、、EI(收录号：20195107857189)、北大核心2017、Scopus(收录号：2-s2.0-85076392114)、CSCD2019_2020、北大核心、CSCD

基金：浙江省自然科学基金(LY19F020015);绍兴市公益性技术应用研究计划项目(2018C10015);国家自然科学基金(61871289,61601200);上海市自然科学基金(18ZR1440600,19ZR1461100)资助~~

语种：中文

中文关键词：高效视频编码;数字音视频编解码技术标准;屏幕混合内容;串匹配;通用串预测;受限串模式

外文关键词：high efficiency video coding;audio video coding standard;screen and mixed content;string matching;universal string prediction;constrained string mode

中文摘要：针对屏幕与混合内容视频具有多样性的特点,在我国第二代数字音视频编解码技术标准的屏幕与混合内容视频编码(the second-generation Audio Video Coding Standard Screen and Mixed Content Coding, AVS2-SMCC)扩展版中,提出了由一般串模式、第一受限串模式和第二受限串模式三种受限串模式组成的通用串预测算法(Universal String Prediction, USP)及其关键技术,其主要部分已经被AVS2-SMCC扩展版工作草案3.0稿采纳.三种受限串模式统一采用偏移串、坐标串和不可预测像素串三种串类型之一或其组合来实现.当编码单元采用USP算法进行编码时,用三种受限串模式对其分别进行预编码,自适应采用三种受限串模式中取得率失真值最小的受限串模式对其进行编码.实验结果表明,对于AVS2-SMCC通用测试序列中移动的文字和图形类别,提出的USP算法与国际上最新的(High Efficiency Video Coding, HEVC)标准SMCC扩展版(High Efficiency Video Coding-Screen Content Coding, HEVC-SCC)相比,在编解码复杂度增加较少的情况下,对于有损全帧内配置 Y 分量、 U 分量、 V 分量平均BD - rate降低率分别可达23.2%、18.3%、19.4%,基于CPSNR的三分量综合的平均BD - rate降低率为22.1%,编码效率已经大幅超过HEVC-SCC.

外文摘要：Screen content coding (SCC) is a key technology for many popular applications in traditional and emerging market sectors, such as cloud computing, remote desktop, screen sharing, etc. Typical computer screen content has very different characteristics from traditional camera-captured content. Therefore, in recent years, SCC has become a hot topic in multimedia applications and has attracted increasing researcher attention from both academia and industry. Two international video coding standards include efficient SCC capability. One is High Efficiency Video Coding (HEVC), a joint effort of the ISO/IEC MPEG and ITU - T SG16 WP3 Q6. The other is the second-generation AVS (AVS2), a joint effort of the Audio Video Coding Standard (AVS) Workgroup of China and the IEEE. Repeated identical patterns (i.e., matching patterns) are often observed on the same picture of screen content. Two major SCC tools in HEVC SCC developed in recent years to exploit those repeated identical patterns with a variety of sizes and/or shapes are: Intra Block Copy (IBC), palette coding (PLT). IBC is efficient for coding repeated identical patterns with a few fixed sizes and shapes. PLT can code repeated identical patterns of two simple cases called LEFT run and ABOVE run inside a CU using two modes (LEFT and ABOVE) of intra-CU. Beyond the HEVC-SCC extension, another tool String Matching(SM) can code general matching patterns with a variety of sizes, shapes, and positions efficiently. Each of the three tools plays an indispensable role in coding some types of screen content. AVS2 is the second- generation video coding standard developed by the Audio and Video Coding Standard (AVS) Working Group of China. Compared to HEVC, AVS2 can achieve significant coding efficiency improvement for scene video coding by adopting intelligent coding tools. In recent years, AVS is developing an AVS2 Screen and Mixed Content Coding (SMCC) extension (AVS2-SMCC). In the early stage of the development, a pixel string matching technique was proposed for AVS2 SMCC extension. Although the technique provided significant coding gain for screen content, it was still less efficient than HEVC SCC extensions for some screen and mixed contents. To fully exploit both local and non-local, both general and special, and both complex and simple matching patterns with a variety of sizes and/or shapes and/or positions in a wide range of commonly seen screen content, a universal string prediction (USP) approach and its key technologies with three modes: general string (GS) mode, constrained string 1(CS1)mode, and constrained string 2 (CS2) mode are proposed, which have been adopted by the AVS2-SMCC extension draft version. The three constrained string modes are implemented with one of the three types of strings: offset string, coordinate string, and unpredictable pixel, or its combination of them. When using USP algorithm to code a coding unit, one of the three constrained modes which gets the minimum rate distortion value is selected to code the CU. Compared with the latest HEVC-SCC extension, experimental results show that, for text and graphics with motion of AVS2-SMCC test sequences, the proposed algorithm achieves the average BD - rate reduction of 23.2%, 18.3% and 19.4% for Y, U and V, respectively in All Intra configuration at about the same degree of the encoding and decoding complexity percentage increase.