高效视频编码帧间预测中运动估计算法研究

发布时间：2018-02-21 09:30

本文关键词： HEVC 帧间预测运动估计搜索策略匹配准则　出处：《重庆大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着信息技术的不断革命和多媒体技术的不断创新，H.264/AVC视频编码标准已经不再能够满足当前社会需求。2013年1月26日，新标准—高效视频编码HEVC(High Efficiency Video Coding)正式问世，它是由两个组织共同制定的，即国际电信联盟ITU-T（International Telecommunication Union）的视频编码专家组VCEG（Video Coding Experts Group）、国际标准化组织ISO/IEC（InternationalStandard Organization/International Electro Technical Commission）的运动图像专家组MPEG（Moving Picture Experts Group），它能获得更加优越的压缩性能，但却需要更多的编解码时间。运动估计作为视频编码的核心技术，耗费时间最多，，因此，对运动估计的搜索策略和匹配准则这两大关键因素进行深入研究是非常有必要的。论文首先对HEVC标准原理、新技术、新概念进行了详细阐述，通过实验对比了H.264/AVC标准，验证了HEVC标准对压缩性能的提高和对高清甚至超高清视频的优势。然后介绍了HEVC标准帧间预测的相关原理，提出了一种基于预测流程的差别判别划分方法，缩短了0.7%以上的编码时间。最后从搜索策略和匹配准则两个方面对块匹配运动估计算法进行研究。在搜索策略方面，分析了常用的运动估计搜索策略的基本原理，并通过实验将其对比了HEVC标准中混合搜索策略；对于匹配准则方面，在研究常用匹配准则的基础上，提出了一种稀疏统计二次Renyi熵的运动估计匹配准则，该准则在计算二次Renyi熵时，引入了统计直方图，采用统计的方式来计算概率密度函数，并结合基于梯度的图像质量评价和运动矢量中心偏离特性，对直方图的统计区间进行稀疏化。通过实验仿真，证明了本文所提出的匹配准则对运动剧烈的视频序列能够得到优于传统绝对误差函数的PSNR值，取得更好的效果。
[Abstract]:With the continuous revolution of information technology and the continuous innovation of multimedia technology, the H.264% AVC video coding standard is no longer able to meet the current social needs. In January 26th 2013, a new standard, HEVC(High Efficiency Video coding, was formally introduced. It was developed by two organizations, VCEG(Video Coding Experts Group, a video coding expert group of the ITU-T(International Telecommunication Union, and MPEG(Moving Picture Experts Group, an expert group on motion images from the ISO/IEC(InternationalStandard Organization/International Electro Technical, which offers better compression performance. As the core technology of video coding, motion estimation takes the most time, so it is necessary to study the two key factors of motion estimation: search strategy and matching criterion. In this paper, the principle, new technology and new concept of HEVC standard are described in detail, and the H.264 / AVC standard is compared through experiments. The advantages of HEVC standard in compression performance and high definition or even ultra high definition video are verified. Then, the relative principle of frame prediction in HEVC standard is introduced, and a differential discrimination method based on prediction flow is proposed. Finally, the block matching motion estimation algorithm is studied from two aspects: search strategy and matching criterion. In the aspect of search strategy, the basic principle of common motion estimation search strategy is analyzed. The hybrid search strategy in HEVC standard is compared by experiments, and a sparse statistical quadratic Renyi entropy matching criterion for motion estimation is proposed on the basis of studying common matching criteria. When the quadratic Renyi entropy is calculated, the statistical histogram is introduced, the probability density function is calculated by statistical method, and the image quality evaluation based on gradient and the center deviation of motion vector are combined. The statistical interval of histogram is sparse and the experimental simulation shows that the proposed matching criterion can obtain the PSNR value of the traditional absolute error function better than that of the traditional absolute error function.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【参考文献】