项目名称: HEVC的低复杂度和并行编码方法研究
项目编号: No.61301116
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 无线电电子学、电信技术
项目作者: 张小云
作者单位: 上海交通大学
项目金额: 24万元
中文摘要: 针对新一代视频压缩标准HEVC面临的更高复杂度、超额运算量、巨大数据量问题,研究面向超高清视频的HEVC低复杂度编码方法:(1)深入分析HEVC编码复杂度,对CU判决、编码模式选择、插值和代价函数计算等复杂度高的模块进行研究,提出相应的快速方法;(2)分析研究超高清视频的统计特性,进一步优化模式判别、阈值判定、提前终止等方法,使得编码运算量不随着超高清视频数据量的增加而成比例地增加;(3)分析各编码模块之间的关系和相互影响,对各个模块进行组合优化,提出一套系统优化的HEVC低复杂度编码方法,在率失真性能下降有限的情况下,大幅度降低编码复杂度。针对面向多核处理器的视频编码器面临的并行设计难点,深入分析视频编码器的功能和数据依赖关系,识别编码器并行执行时存在的各种相关性,并结合HEVC的Tile和WPP等并行特征,建立正确高效的并行任务分解模型,提出适用于多核处理器的HEVC并行编码方法。
中文关键词: HEVC;视频编码;率失真优化;模式判决;并行
英文摘要: The new generation of video compression standard HEVC faces the challenges of higher coding complexity, huge amount of computation, great amount of data for ultra-high-definition video, and parallel implementation on multi-core processors. To solve the above problems, we focus on low complexity and parallel coding methods for ultra-high-definition video based on HEVC. First, we fully analyze the complexity of HEVC coding and find out the modules which have high complexity and low encoding performance, and propose fast and efficient algorithms. Then, we explore the statistical properties of the ultra-high-definition video, and develop HEVC low complexity coding methods by making use of such properties. Also, we analyze the interaction between these various coding modules, integrate these fast coding methods together and put forward a system-level optimized HEVC encoder with low complexity. As for the difficulties in designing efficient parallel computing for video coding based on multi-core processor, we study the task and data dependency, task parallelization and scheduling, and propose the efficient parallelization method for HEVC encoder.
英文关键词: HEVC;Video Coding;Rate Distortion Optimation;Mode Decision;Parallel