【作者】 黄剑飞；

【导师】 高文；

【作者基本信息】 中国科学院研究生院 ， 计算机应用技术， 2008， 硕士

【摘要】 随着互联网和数字多媒体技术的迅猛发展,数字视频的应用越来越广泛。经过多年的发展,视频编码技术已经较为成熟,最新的视频编码国际标准H.264/AVC以及我国的AVS在编码效率上都比以往的标准提高了一倍以上。然而,随着技术的持续进步,从VCD到DVD,从标清到高清,人们对视频质量的要求也随之提高。因此,对视频编码中的关键技术进行深入研究,不断进行优化,是非常具有理论和现实意义的。此外,由于视频编码标准不断的推陈出新,在新旧标准之间的转码成为研究热点。在我国,目前被广泛应用的MPEG-2标准与新一代的AVS标准之间的转换,具有极大的应用价值。本文研究视频编码中率失真函数与码率控制等若干关键问题,提出了基于H.264标准的率失真模型以及Two-pass VBR码率分配算法;研究MPEG-2/AVS转码中的码率控制问题,提出了平滑质量的码率控制算法。本文研究的主要内容包括:1.基于H.264/AVC标准的率失真函数分析本文首先利用信息论和量化理论的知识,着重研究如何应用码率失真理论分析实际编码的码率失真曲线(也称码率失真函数、率失真函数),得到基于H.264/AVC编码标准的率失真模型。具体使用广义高斯分布GGD来描述编码中变换系数的分布,并在H.264/AVC特定编码量化模式下进行推导,最终得到其率失真模型,为平滑的视频质量控制打下基础。2.基于率失真模型的码率分配算法在得到率失真模型之后,本文进一步研究如何利用该模型,实现保证质量的码率控制方法。为此,本文基于H.264/AVC编码标准,开发了一种全新的Two-pass VBR编码算法。该方法首先通过第一次编码确定视频序列中每一帧图像的率失真函数模型,并利用该模型计算目标码率,再根据目标码率调整量化系数,从而在第二次编码中得到质量恒定的编码结果。实验表明,和原有编码算法相比,这种新方法能够在保证目标码率的前提下提供更稳定的视频质量。3. MPEG-2/AVS视频转码中的码率控制算法本文还对MPEG-2/AVS视频转码中的码率控制进行了一定研究,得到了一种保证质量平滑的转码方法。首先,利用一种复杂度率失真模型对码流进行分析,根据目标码率计算初始量化系数。接下来,在转码过程中根据原始MPEG-2序列,使用一定的控制算法动态调整量化系数,以达到合理的目标码率。实验表明,本算法能够在保证目标码率的前提下得到较为平滑的视频转码质量,提供更好的主观质量感受。这些视频编码与转码的问题既有一定的理论深度和学术价值,同时又对应用有直接的指导意义。更多还原

【Abstract】 Over the past few decades, digital video compression technique was extensively used in video communication, TV broadcast, video on demand, and digital storage. Several international industry standards have been established, including MPEG-1/2/4, H.263 and H.264. The newest H.264/AVC and AVS video coding standards doubled the coding efficiency relative to the performance of prior standards such as MPEG-2.For all video encoders, rate control plays an important role. It directly affects the bit rate and quality of the output bit streams. So it is very important for us to have an in-depth study on the key techniques used in rate control and improve the coding efficiency of the new coding schemes. In this paper, we present a two-pass variable bit rate (VBR) coding algorithm for H.264/AVC, using a novel R-D model. The model is developed based on the statistical analysis of the integer transform coefficients with generalized Gaussian distributions (GGD). Also, we develop a new VBR bit allocation method for MPEG-2/AVS video transcoding. The main contributions in this paper are listed as follows:1. Rate-distortion analysis and modeling of the H.264/AVC standardIn this paper, we first present a new rate-distortion model for H.264. This model is developed from the statistical analysis of integer transform coefficients using GGD. The analysis and modeling are based on information theory and the quantization scheme of H.264.2. A novel Two-pass VBR bit allocation algorithm for H.264/AVC standardIn this paper an effective VBR algorithm for H.264/AVC is developed, which adopts two-pass encoding to achieve a constant video quality. Characteristics of the entire video sequence are collected and analyzed in the first pass, using our new R-D model, and then the sequence could be optimally re-encoded in the second pass. Experimental results show that the proposed algorithm is effective and efficient to reduce the PSNR variation and provide consistent visual quality.3. A new rate control method for MPEG-2/AVS transcodingIn this paper, a simple and effective rate control method for MPEG-2/AVS transcoding is presented. First, the initial quantization parameter of the transcoding is achieved by a new R-D model in transcoding. Then, during the AVS encoding, quantization parameter is adjusted according to the MPEG-2 bit rate using our new rate control algorithm. Experimental results show that the proposed algorithm can effectively reduce the PSNR variation and provide consistent visual quality.更多还原