【作者】 李志刚；

【导师】 张兆扬；

【作者基本信息】 上海大学 ， 通信与信息系统， 2005， 博士

【摘要】 H.264/AVC是ITU-T和ISO联合推出的新标准,采用和发展了近几年视频编码方面的先进技术,以较高编码效率和网络友好性而著称。它仍基于以前视频编码标准的运动补偿混合编码方案,主要不同有:增强的运动预测能力;准确匹配的较小块变换;自适应环内滤波器;增强的熵编码。测试结果表明这些新特征使编码效率比以前的标准约提高50%,但增加了复杂度。本文着重研究经传输层的若干应用问题。随着新的视频编码标准H.264/AVC的逐步推广应用,如何将H.264/AVC视频流经MPEG-2的系统传输层传输,对于充分利用原有大量的MPEG-2系统有重要意义。对此提出一种解决方案,要点是:先把H.264/AVC视频作为MPEG-2系统层传输的基本流,然后扩展MPEG-2标准中的传输流系统目标解码器(T-STD),使之可以将H.264/AVC编码视频复用到MPEG-2传输流(TS)包中在Internet上传输和解码。被解码的基本流通常来自于一个“容器”(如AVI或者TS),在客户端从服务器端的这个容器中取出H.264/AVC基本流后便可实时解码、显示。经仿真实验表明,该方案能够获得较好的流视频效果,在带宽受限的情况下信噪比低于40dB的帧数少于5%,可用于IP网络流视频或移动视频中。在完成H.264算法优化和多媒体指令集优化之后,将其扩展至基于比特平面编码的FGS编码,从而实现了基于PC平台的实时SNR(信噪比)精细粒度可伸缩编码,CIF格式编码帧速可达30fps以上。此外,为使H.264/AVC码流能在带宽较大变化的IP和无线等异构网络中传输,本文另外提出一种将H.264扩展至混合空域/时域/SNR精细可伸缩方案,并给出了相应的码率控制策略:根据率失真优化的结果来选择QP(量化参数),在编码器端对基本层作GOP(图像组)级的码率控制,而对增强层作逐次精细化的码率控制。本文方案与由JM8.6的基本层码率控制、FGS比特平面截断作增强层码率控制(简称JM8.6+FGS)的方法相比,视频质量更高,PSNR(峰值信噪比)变化更为平滑。且在接收端,在某一目标比特率约束下,可以由不同时间分辨率(帧速率)和不同空间分辨率(图像格式)配置的终端截取和实时解码。仿真结果表明,本文方法的亮度平均峰值信噪比(Y-PSNR)在CIF格式时优于JM8.6+FGS方法达2.45dB,且与目标比特率更为匹配;也比新近提出的JVT-N020提案,在平均Y-PSNR上有0.15dB的增益,而且图像质量更为平滑。H.264/AVC支持SP(同步预测)帧,允许不同质量比特流之间的高效切换,MPEG-4支持FGS编码。本文提出一种将两者融合在一起的解决方案,在JM联合模型中实现了流切换的功能,并且加入FGS编码,使得传输的比特流既能适应因特网或无线网传输带宽的大跨度波动,又能灵活适应小范围的带宽变化。仿真实验结果表明:本文所提出方案的亮度Y分量峰值信噪比比FGS平均好0.47dB,比流切换方法平均好0.23dB。且在专为移动业务而设计的H.264/AVC更多还原

【Abstract】 H.264/AVC is a new video codec standard accepted by ITU-T and ISO, which adopts many advanced technologies and develops them in video codec since recent years, and becomes famous for its higher coding efficiency and better network friendship than others. Based on the motion-compensated hybrid coding scheme originated from previous video codec standards, H.264/AVC has other important differences as follows: the enhanced motion prediction capability; accurate matching for small block-size transforming; adaptive in-loop deblocking filter; enhanced entropy coding. The experimental results are shown that these new characteristics promote the encoder efficiency by about 50% than the previous counterparts, but the complexity addition is inevitable. This thesis focuses all our attentions on some applications studies throughout the transport layer in the IP networks.With the increasingly extensive applications for the new emerging video coding standard, H.264/AVC, it becomes more and more important that we transport H.264/AVC video stream over MPEG-2 system by exploiting all the existing large amount of the infrastructure of MPEG-2 systems. We propose a solution for it. The key technologies are as follows: Firstly we use an H.264/AVC video as an elementary stream of MPEG-2 system, then extend the Transport stream-System Target Decoder(T-STD) of MPEG-2 standard, so as to pack the H.264/AVC video elementary stream into MPEG-2 Transport Stream(TS) to transport through Internet and decode it in the client. The decoded elementary stream should ordinarily come from a container, such as AVI or TS. We extract the H.264/AVC video from this container to be real-time decoded and be represented in the client after the reception. The experiment results show that we get a good effect, with less than 5% frames whose PSNR<40dB even in the case of bandwidth-constrained scenario. Running in the IP networks, the streaming solution can also be used for mobile video.After the algorithms optimization and multi-media instruction-set optimization for the H.264 encoder, the FGS (Fine-Granular Scalability) based on bit-plane coding is extended to realize a real-time SNR fine-granular scalable coding based on PC platform, and the frame rate can be 30 fps and over for the video with CIF format. Moreover, to enable transmission of H.264/AVC bit-stream over IP and wireless heterogeneous networks with randomly variable bandwidths, another scheme of hybrid spatial/temporal/SNR refined scalability and its rate control strategy are given: the rate control for GOP (Group of Pictures) level is done on the base-layer at the encoder side to select the QP (Quantization Parameter) according to the RDO (Rate Distortion Optimization) results, and the progressively refined rate control is done on the enhancement-layer. In comparison with the method of JM8.6+FGS (i.e., the base-layer rate control is based on JM8.6, and the enhancement-layer control is更多还原