新一代视频编码标准HEVC参考帧设置与块划分快速算法研究

发布时间：2018-03-21 23:27

本文选题：高效视频编码　切入点：HEVC　出处：《华南理工大学》2014年硕士论文　论文类型：学位论文

【摘要】：2013年3月，ITU-T的视频编码专家组(VCEG)和ISO/IEC的运动图像编码专家组(MPEG)联合开发的下一代视频编码标准HEVC（High Efficiency Video Coding）正式发布。作为H.264/MPEG-4Part10AVC的后继者，HEVC同样基于传统的混合视频编码框架。HEVC采用了许多先进的技术，包括基于四叉树的块划分、高达32x32大小的变换单元、更加精细的帧内预测模式、全新的运动融合技术、高处理效率的上下文自适应二进制算术编码等等。这些技术的使用使得HEVC的压缩性能提高了将近一倍，并特别适用于高分辨率（如4K、1080p）视频的压缩。 HEVC压缩性能提高的同时，其计算复杂度也成倍增长，这很大程度上影响了标准的推广。为了减少编码时间，本文主要的研究工作主要从以下两个方面展开。一是多参考帧的设置问题，通过对HM中低延时配置下参考帧的分布进行统计，发现对于高QP值2幅图像而言，97%以上区域的参考指向集中在前2个和前3个参考帧内。为此，本文提出了分别舍弃高QP值图像2个和1个参考帧的简化方案。测试结果表明，在使用HM中已有快速算法的基础上，新方案能够继续减少9.5%的编码时间，而相对码率只增加了0.57%。二是编码单元划分问题。因为采用了大小灵活的编码单元技术，编码时间成倍增加。本文深入研究了HM中已有的块划分算法，针对HM中的低延时配置提出了改进的“EarlyTermination”算法。该算法通过分析低QP值图像中不同大小块的率失真代价来获取相应的终止阈值，并提出了简单的阈值更新策略。结果表明，在使用HM中已有快速算法的基础上，该算法能够继续减少12.7%的编码时间，而码率增加了2.4%。本论文内容组织如下：第一章介绍了视频编码的基本原理、历史标准，以及当前HEVC的研究热点等；第二章对HEVC的主要关键技术模块进行了详细阐述，为后文的研究打下基础；第三章在对HEVC中的多参考帧技术进行分析研究的基础上，，针对低延时配置提出了简化的参考帧设置方案，并对优化方案进行了测试；第四章通过对HM中编码单元划分和模式选择的研究，提出了改进的“EarlyTermination”算法，并对算法进行了测试。
[Abstract]:In March 2013, the next generation video coding standard, HEVC(High Efficiency Video coding, which was jointly developed by ITU-T video coding expert group and ISO/IEC moving image coding expert group, was officially released. As a successor to H.264/MPEG-4Part10AVC, HEVC(High Efficiency Video coding is also based on traditional hybrid video coding. The code frame. HEVC uses many advanced technologies, Including block partitioning based on quadtree, transform units up to the size of 32x32, more refined intra prediction mode, new motion fusion technology, Context-adaptive binary arithmetic coding with high processing efficiency and so on. The use of these techniques makes the compression performance of HEVC nearly double, and is especially suitable for high resolution video compression (such as 4KT 1080p). With the improvement of HEVC compression performance, the computational complexity increases exponentially, which greatly affects the popularization of the standard. The main research work in this paper is as follows: the first is the setting of multi-reference frames, which is based on the statistics of the distribution of reference frames in HM low-delay configuration. It is found that for two images with high QP value, the reference points of more than 97% regions are concentrated in the first 2 and the first 3 reference frames. In this paper, a simplified scheme of abandoning 2 and 1 reference frames of high QP value images is proposed respectively. The test results show that, On the basis of using the fast algorithm in HM, the new scheme can reduce the coding time by 9.5%, but the relative code rate can only increase by 0.57%. The second problem is the partition of coding units. Because the coding time is multiplied by using the technique of flexible coding units, the block partitioning algorithms in HM are deeply studied in this paper. This paper proposes an improved "EarlyTermination" algorithm for low delay configuration in HM. The algorithm obtains the corresponding termination threshold by analyzing the rate-distortion cost of different size blocks in the low QP-valued image, and proposes a simple threshold updating strategy. On the basis of using the fast algorithm in HM, the algorithm can reduce the coding time by 12.7% and increase the bit rate by 2.4%. The main contents of this thesis are as follows: the first chapter introduces the basic principle of video coding, the historical standard, and the current research hotspot of HEVC, the second chapter describes the main key technology modules of HEVC in detail, which lays the foundation for the later research. In chapter 3, based on the analysis and research of multi-reference frame technology in HEVC, a simplified reference frame setting scheme is proposed for low-delay configuration, and the optimization scheme is tested. In chapter 4, an improved "EarlyTermination" algorithm is proposed by studying the partition of coding units and mode selection in HM, and the algorithm is tested.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】