【作者】 王正宁；

【导师】 诸昌钤；

【作者基本信息】 西南交通大学 ， 计算机应用技术， 2007， 博士

【摘要】 H．264/AVC是ITU-T VCEG和ISO/IEC MPEG联合制定的最新国际视频编解码标准。标准以较高的编码效率和网络友好性受到人们的广泛关注。虽然H．264/AVC采用的依然是基于块的混合视频编码框架，但由于集中了近些年来视频编解码领域内的一系列最新技术成果，在同等图像质量条件下其压缩率较以往标准提高一倍以上。所采用的新技术主要包括：空间域内的多种模式帧内预测，1/4像素精度运动估计，整数DCT变换量化，多参考帧运动补偿，自适应去块效应滤波以及基于内容的熵编码等。采用新的编码技术虽然在编码效率上得到了很大的提高，但与以往视频编解码标准相比，H．264/AVC的计算复杂度却高出几倍甚至十几倍。非常高的计算复杂度严重影响了H．264/AVC在实时编码方面的应用和推广，因此如何在保证编码效率的前提下尽可能降低编码计算复杂度成了人们研究的热点。本文针对H．264/AVC编码部分计算量较大的三个部分：帧内预测，帧间块划分模式选择和多参考帧运动估计，进行了相关快速算法的研究。首先，本文在深入分析了H．264/AVC帧内预测机制及编码模式选择算法的基础上，提出了一种基于频域特征的快速帧内编码模式预测算法来有效降低帧内预测的计算复杂度。算法首先对编码块进行一次DCT变换得到所需的一些DCT系数，然后根据这些系数计算出编码块的纹理方向和纹理方差。接着通过计算得到的纹理方向和方差对当前编码块的帧内编码模式进行有效地预测。实验结果表明本文算法在保证编码效率与标准算法几乎一致的前提下将帧内预测的计算复杂度平均降低了67．5％，而且性能方面优于同类算法。在基于频域特征的帧内编码模式快速预测算法的有效性得到验证之后，接着又从提高预测准确性和增强硬件实现方面模块的复用性两方面出发，以前面本文所提算法为基础提出了考虑相邻像素点的5×5 DCT和复用H．264/AVC整数DCT变换的两种扩展算法，相应的实验结果也分别验证了两种扩展算法的有效性。其次，本文仔细研究了H．264/AVC标准中帧间编码时块划分模式的选择算法并通过大量实验分析了实际编码中块的划分模式选择特点，提出了一种基于残差纹理分析的快速块划分模式选择算法。算法首先通过一次宏块级的运动搜索预处理得到各个4×4块运动补偿后的残差：然后根据得到的残差纹理大小及分布情况对块的帧间划分模式做出快速选择。大量的实验验证了本算法在几乎不牺牲图像质量和压缩效率的基础上将这一过程的计算复杂度平均减少了59．85％。与相同时期的同类算法相比，在计算复杂度的降低程度、图像质量以及码率变化方面均优于这些算法。最后，本文在分析了H．264/AVC多参考帧条件下帧间预测编码算法特点的基础上，从如何有效减少运动估计过程中参考帧的使用数量以及如何加快各参考帧中运动估计的计算速度两个角度出发，先对影响H．264/AVC多参考帧预测编码的一些因素进行分析，然后以视频序列中物体运动的连续性为基础给出一种多参考帧条件下的运动向量预测方法，并以全零块预判技术制定出相应的运动搜索停止策略，最后综合相关的因素和策略提出了一种多参考帧条件下的快速运动估计算法。算法在不同实验参数配置下对多个标准测试序列进行编码，实验结果表明本文算法以编码效率方面较小的牺牲将这一过程的计算复杂度平均降低了63．1％。更多还原

【Abstract】 H.264/AVC is the newest international video coding standard jointly developed by the ITU-T VCEG and the ISO/IEC MPEG standards committees. It represents the state-of-the-art video compressing technology, and addresses the full range of video applications including low bit-rate wireless video applications, standard-definition and high-definition broadcast television, and streaming over the Intemet. Compared with MPEG-4, H.263, and MPEG-2, H.264/AVC can achieve 39%, 49%, and 64% of bit-rate reduction with equivalent video quality respectively. To achieve such a high coding efficiency, H.264/AVC adopts many new and advanced features such as spatial intra prediction, quarter-pixel accuracy motion compensation, integer 4×4 discrete cosine transform and quantization, variable block sizes, multiple reference frames, adaptive de-blocking filter and context-based entropy coding.As might be expected, the increase in coding efficiency and coding flexibility comes at the expense of several times in complexity with respect to earlier standards. The high coding complexity has significant impact on the applications of real time video coding such as video conference and telephony. In order to reduce the computational complexity in H.264/AVC effectively, many fast algorithms have been developed. In this paper, intra prediction, inter partition mode selection and motion estimation with multiple reference frames are addressed.First, a fast H.264/AVC intra prediction mode selection algorithm is proposed in this work. The objective is to reduce the encoder complexity without significant rate-distortion performance degradation. The proposed method uses the frequency domain features of the target block to filter out the majority of candidate modes. By adopting DCT operation on each block, the algorithm uses the energy ratio and variance to predict the probable modes for current block. Extensive simulations verify that the proposed method speed up the intra prediction process by 67.5% on average without sacrificing the picture quality and compression ratio.When the fast frequency domain features based algorithm has been proved to be effective, its two extended algorithms are proposed. One considers the neighboring pixels to get a more accurate prediction, and the other one uses the integer DCT defined in H.264/AVC to enhance the reusability in hardware implementation. It is demonstrated by the simulation results that the two extended methods are robust and effective.Second, a residual texture based fast block size selection algorithm for inter-frame coding is proposed in this paper. Firstly, motion estimation (ME) is executed for current macroblock and the residual is computed; then the block size is predicted by analyzing the residual texture. Experiment results verify that the proposed method speeds up the block-size selection procedure by 59.85% without sacrificing picture quality and compression ratio. A comparison of the proposed algorithm with other algorithms is also provided.Third, a novel scheme to reduce the complexity of multiple reference frame motion estimation in H.264/AVC is proposed. As we know, the computational complexity of multi-reference frames motion estimation is increased in proportion to the number of searched reference frames. In order to reduce the complexity of multiple reference frames motion estimation in H.264/AVC effectively, the following factors such as frame rate, motion activity and search range are considered by the proposed algorithm. Using the motion continuity, a multiple reference frame motion vector prediction scheme is described in the proposed method. Furthermore, an all zero block detection method is used to speed up the motion estimation process. Experiment results show that the proposed algorithm can reduce the number of reference frame from 5 to 1.7 and save the multiple reference frame motion estimation time by 63.1% on average respectively with little degradation of coding efficiency.更多还原