【作者】 滕国伟；

【导师】 张兆扬； 张一钧；

【作者基本信息】 上海大学 ， 通信与信息系统， 2005， 博士

【摘要】 在过去的十几年中,国际标准化组织制订了一系列的国际视频编码标准并广泛应用到各种领域。其中,ITU-T制订了满足实时视频通信应用的标准,例如H.261、H.263、H.263+和H.263++;ISO制定了满足视频存储、广播视频和视频流需要的MPEG标准,包括MPEG-1、MPEG-2及MPEG-4等。而H.264/AVC是ITU-T和ISO联合推出的新标准,采用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成为新一代国际视频编码标准。 H.264/AVC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H.264/AVC编码器提高50%编码效率的同时,增加了一个数量级的复杂度。实际中恰当地使用H.264/AVC编码工具可以较低的实现复杂度性得到与复杂配置相同的编码效率。故实际编码系统开发需要在运算复杂性和编码效率之间进行折衷,兼顾考虑。 H.264/AVC引入的新编码特征既增加基本模块的复杂度,也成倍增加算法的复杂度。针对它们的作用和实现方法的不同,可采用不同的优化方法:对于基本模块可用多媒体指令来优化;对于实现算法可研究快速算法来代替。本文基于上述思路进行优化,具体的工作包括: 针对帧内预测编码的复杂性,本文提出一种基于率失真优化(RDO:Rate Distortion Optimization)的快速帧内预测模式选择算法。该算法利用当前宏块色度块和亮度块帧内预测模式间的相关性、4x4亮度块和16x16亮度块模式选择之间的相关性以及亮度块预测模式内部的相关性,并以RDO为主,结合较简单的SATD(Sum of Absolute Transform Difference)为判决依据快速进行帧内预测模式选择。 针对帧间编码模式的复杂性,本文提出一种快速模式判决算法。该算法将全部编码模式进行分类,对SKIP模式采用提前终止判决,而对其它模式可通过判别宏块的一致性和比较宏块的空时相关性逐类型比较,可有效减少判决的模式数量。 针对运动估计的复杂性,结合H.264/AVC运动估计引入的新特征,本文提出了一种快速整像素运动估计算法。该算法通过统计和分析当前块的相关矢量确定运动估计的搜索范围和方向,再辅以有效搜索路径,可以显著减少搜索点数量并能达到较好的匹配效果。 针对多参考帧运动估计的复杂性,本文提出一种快速多参考帧选择算法。该算法根据实际中多参考帧的应用特点,利用序列的纹理特征来判定可能的参考帧数目,然后根据运动信息和RDO的单调性进一步确定16x16,8x8和4x4模式的最佳参考帧,而其它模式可根据与这三种模式的参考帧选择相关性来自适应确定可能选中的参考帧,有效降低搜索帧数目,避免搜索不必要的参考帧。 本文基于上述算法还进行Baseline Profile编码器的研究,以简化RDO为模式判决依据,给更多还原

【Abstract】 In the past about ten years, a series of international video coding standards are developed and widely used in a variety of domains. Some standards for real-time telecommunication applications have been established by ITU-T, including H.261, H.263 and later enhancements of H.263 known as H.263+ and H.263++, Others for video storage, broadcast and stream applications by ISO/IEC, such as MPEG-1/2/4. H.264/AVC is the newest standard approved by the ITU-T and ISO/IEC organizations, which represents a number of advances in video coding technology and becomes the newest international video coding standard because of both high coding efficiency and network friendliness.The design of H.264/AVC is based on conventional block-based motion-compensated hybrid video coding concepts, but with some important differences relative to prior standards: enhanced motion-prediction capability; a small block-size exact-match transform; adaptive in-loop deblocking filter; enhanced entropy coding methods. The experiment results show that the improved coding efficiency, up to 50%, comes with a complexity increase of more than one order of magnitude at the encoder. In fact, a proper use of the H.264/AVC tools leads to roughly the same performances as the complex configuration. So both computational complexity and coding efficiency should be thought and made to the optimal trade-off in the implementation of actual coding system.The new features of H.264/AVC increase not only the complexity of coding basic modules, but also the one of algorithms by times. The multimedia instruction may be used to optimize the basic modules; the fast algorithms are developed instead of the old ones. This dissertation is based on the above methods, and major works are as follows.To reduce the complexity of intra prediction, a fast intra-prediction mode selection algorithm is proposed. It uses the mode selection correlation between chroma block and luma block, the correlation between luma 4x4 block and luma 16x16 block and the selection correlation in luma block modes with different criterions(RDO(Rate Distortion Optimization) or SATD(Sum of Absolute Transform Difference)) to fast determine the best mode.To reduce the complexity of inter prediction, a fast inter mode decision algorithm is proposed. It classifies all modes by both homogeneity and space-time correlation of macroblock. The SKIP mode can be early terminated, and other modes can be compared type by type. So the number of calculated modes can be decreased efficiently.To reduce the complexity of motion estimation, a fast integer pixel motion estimation algorithm combined with the new features introduced by H.264/AVC is proposed. It analyzes relevant motion vectors of current block to determine search range and direction, and designs effective search path, which can efficiently decrease the number of search points and come to a good matching.To reduce the complexity of multiple reference frames, a fast selection algorithm of reference更多还原