给予HVS的H.264/AVC码率控制算法研究

	给予HVS的H.264/AVC码率控制算法研究
	孙乐
学位类型	博士
导师	戴明
	2014-07
学位授予单位	中国科学院大学
学位专业	光学工程
摘要	二十一世纪，随着多媒体技术的发展，视频通信以及网络传输技术得到了广泛地应用。视频作为一种内容丰富且直观的信息载体，正对人们的交流和社会发展起着越来越重要的作用。首先，当今社会，视频通信正在逐步取代原有的单一语音和文字通信，手机3G视频通信、远程视频会议、远程监控系统等前沿技术已逐步成为信息时代的沟通利器和保证社会稳定的强有力武器。其次，在个人娱乐领域，爆炸式的视频信息在人们的日常生活中占据着主导地位，人们对社会信息获取的方式也从单一的文字向视频信息转变，如：优酷、Youtube等网络视频平台。随着后PC时代的到来以及人类对移动终端的追捧，移动终端的视频需求与日俱增。伴随着信息技术的不断发展，消费者对视频体验也逐步提高，也因此，视频通信也将面临前所未有的机遇和挑战。视频编解码技术是保证有效传输和存储视频信息的关键技术之一，也是现在前沿信息技术不可或缺的重要组成部分。H.264/AVC（以下简称H.264）是ITU和ISO/IEC联合制定的视频编码标准，也是MPEG-4的第10部分。2013年3月，ITU和ISO/IEC公布了最新的视频编解码标准，即HEVC（High Efficiency Video Coding）。HEVC是针对高清视频格式的编码标准，基本框架是基于H.264的标准框架，并对视频编解码的各个环节做了进一步的改进以符合高清视频的要求。作为主流的视频编码标准，H.264/AVC引入了许多先进的编码技术，相比以往的编码算法,在编码效率、网络适应力以及压缩质量等性能指标方面都有显著地提高。然而，H.264/AVC引入的率失真优化导致了“蛋鸡悖论”的产生，也因此使得码率控制较之前的编码标准更为复杂。本文针对H.264/AVC编解码标准对码率控制做了大量的分析、实验，最后提出了一种新的基于人眼视觉系统的码率控制方案。本文首先介绍了H.264/AVC编解码器的基本工作原理，如帧内、帧间预测、DCT变换以及量化编码等，并对H.264/AVC标准的基本概念进行了概述。其次介绍了一种H.264/AVC压缩域视频分割算法，该算法利用H.264码流中提取得到的宏块相关运动信息对宏块所属不同区域进行聚类划分。在使用宏块相关运动信息之前，我们对原始运动矢量场、宏块分配的权值场分别进行了可靠性分析以及二维的空间滤波处理，从而得到可信度较高的运动矢量场和宏块的权值场，最终将Meanshift聚类算法应用于可靠的运动矢量场，进而实现运动目标的提取。在恒定信道环境和现有的码率控制方案条件下，为了改善视频的编码质量，本文对JVT-G012提供的码率控制方案（GOP层、帧层、基本单元层的比特分配以及MAD预测模型等）进行大量的实验和分析，并提出了一种新的基于人眼视觉系统的码率控制方案，该方案仍然沿用了JVT-G012码率控制方案的基本框架，但对MAD预测模型、帧层码率分配以及基本单元层的码率分配做了了改进。为了获取较为准确MAD预测值，本文充分利用帧间和帧内MAD的时空相关性提出了一种自适应的时空MAD预测模型；原有的帧层比特分配是通过预设的目标缓冲级实现的，本文对缓冲器占用量的控制参数进行了重新修正，由原来单一的比例环节拓展为比例、积分、微分环节；最后，结合本文提出的区域分割算法，对基本单元层比特分配按照区域视觉特性重新修正。针对传输信道环境受限的应用环境，本文提出的码率控制方案在MAD预测、比特分配的性能方面都有较大的提升，实验结果表明，该码率控制方案有效的提高了重建视频的主观质量。
其他摘要	In the 21st century, with the development of multimedia, the technology of video communication and network transmission has been widely used. As an intuitional and abundant information carrier, the video has been playing an increasingly critical impact in human communication and the social development. First of all, video communication is replacing the single voice and text communications step by step in the current society. The advanced techniques, like the 3G mobile video communication, the remote video conference and the remote video surveillance system etc, have become the very useful communicating approaches and have acted as the powerful “weapons” to ensure the social stabilization. Secondly, as for the personal entertainment, the explosive video information has dominated people’s daily life and people’s need for information has transformed from the single text to the video, for example, Youku and YouTube are the well-operated network video platforms. While the post-pc era is arriving, the pursuit for the mobile terminals and the demand of video which applied at mobile terminals are increasingly large. Moreover, consumers’ expectation for video experience is higher and higher with the support of information technology development. Therefore video communication is currently facing the unprecedented chances and challenges. Video code/decode is the one of the key techniques that ensures information effective transmission and storage and is also the significant part of the cutting-edge information technology. H.264/AVC (hereinafter referred as H.264) is the video coding standard jointly made/recommended by ITU and ISO/IEC and is also the part 10 of the MPEG-4 standard. In March 2013, the ITU and ISO/IEC released the most updated standard i.e. HEVC (High Efficiency Video Coding). HEVC is designed for the high resolution video format, based on the H.264 framework and made the improvements on each part of encode and decode in order to meet the special demand of high resolution. As the mainstream coding standard, H.264/AVC brings in a number of advanced techniques. Compared with the previous encode/decode standards, the aspects on coding efficiency, adaption of network and quality of compression, etc, have been obviously improved. However, the RDO (Rate Distortion Optimization) brought by H.264/AVC causes the chicken-and-egg dilemma. Consequently, the rate control of H.264 becomes more complicated than the previous encode/decode standards. This dissertation proposes a novel rate control algorithm based on the HVS through the experiments and analysis on the standard rate control algorithm (JVT-G012). This dissertation starts with the description of the basic principles of H.264/AVC, such as intra-, inter prediction, DCT transform and quantization coding, and the overview of the basic concepts of H.264/AVC standard.Besides, it introduces a H264/AVC compressed domain video segmentation algorithm which uses the macro-block motion information extracted from H.264 bitstream to cluster the macro-blocks and partition the macro-blocks areas. Before using the macro-block motion information, the reliability analysis and the two-dimension spatial filter processing are respectively carried on the original motion vector field and the value field of macro-block partition, so that a more reliable motion field and macroblock weight field are obtained. Ultimately, mean shift clustering algorithm is carried on the reliable motion field to achieve the moving object extraction. In the constant channel environment and the current existing rate control scheme condition, in order to improve the quality of video coding, this dissertation conducts a number of experiments, analyses and proposes a novel rate control scheme based on the JVT-G012. The proposed rate control scheme inherits the basic framework of JVT-G012 (GOP layer, Frame layer, Basic layer bit allocation and MAD prediction model etc.) and improves the MAD prediction model, the frame layer bit allocation and the Basic unit bit allocation. To obtain the more accurate value of MAD, this dissertation utilizes the spatial and temporal correlation adequately proposing an adaptive temporal-spatial MAD prediction model. In JVT-G012 rate control scheme, the frame bit allocation is achieved by setting a target buffer level. However, in proposed rate control scheme, the parameter is corrected and extended from single proportion to Proportion-Integral-Differential (PID) controller. Eventually, combined with the proposed region segmentation algorithm, the basic unit layer bit allocation is adjusted. For the limited bandwidth application environment, the performance of the proposed rate control scheme is highly improved in MAD prediction and bit allocation. The experiment results show that the proposed rate control scheme for H.264/AVC increases the subjective quality of reconstructed video efficiently.
语种	中文
文献类型	学位论文
条目标识符	http://ir.ciomp.ac.cn/handle/181722/41468
专题	中科院长春光机所知识产出
推荐引用方式 GB/T 7714	孙乐. 给予HVS的H.264/AVC码率控制算法研究[D]. 中国科学院大学,2014.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
孙乐.pdf（4270KB）			开放获取	CC BY-NC-SA	请求全文