【作者】 谢剑锋；

【导师】 宋利；

【作者基本信息】 上海交通大学 ， 信息与通信工程， 2017， 硕士

【摘要】 视频压缩给如今的人们带来了各种生活、工作、娱乐上的便利,比如高清视频、可视电话、视频会议等等,其所依赖的视频编码技术在过去的几十年取得了长足的发展,从早期的H.261到如今的HEVC,视频编码技术通过一代一代的演进大致确立了混合编码的框架结构。在所有的这些编码技术中,码率控制都扮演着举足轻重的角色。在实际应用中,由于带宽资源的有限,要求在限定的码率下保证编码质量的同时,实现输出码率的稳定性,这就是码率控制问题。除了精确的码率控制以外,如何解决码率控制模式下编码质量较低,以及码率控制对并行编码的支持较差的问题,也成为目前视频编码的研究热点。关于如何解决码率控制模式下编码质量较低的问题,本文提出了一种时域依赖的码率控制比特分配方案。相对传统的算法仅仅根据编码单元的内容复杂度来分配比特的方式,时域依赖的比特分配方案除了考虑自身内容复杂度的因素,还考虑到编码单元对后续参考它的编码单元的影响。具体来说,如果某个编码单元在后续帧里面被多次参考,则它在时域上的影响因子较大,应考虑给其分配相对较多的比特,使其以相对较高质量的编码模式编码,从而保证所有与其时域相关的编码单元的编码质量。本文具体描述了该方案的算法过程和实现细节,并通过实验验证了方案的有效性,分析了编码复杂度的影响。在改进码率控制对并行编码的支持方面,本文提出了一种支持多机并行编码的联合码率控制方案,该方案通过独立地分配帧级目标比特和自适应地分配片级目标比特可以达到并行编码下相对准确的码率控制。并行编码框架的流行带来了数据依赖方面的挑战,越高的并行度会导致越高程度上的数据相关性的割裂,这种问题给码率控制带来了巨大的影响,因为码率控制正是基于视频数据在时间以及空间上的相关性,在编码过程中充分地利用已编码帧的历史数据不断地调整编码参数以达到全局的码率稳定性。本文提出的联合码率控制方案通过定期地收集各个独立编码器的编码信息,综合分析后根据各个编码器的实时编码状态自适应地设定合适的编码参数,从而使得合成后的码率能够相对稳定地维持在目标码率。更多还原

【Abstract】 Video compressing brings various conveniences to our daily life,work and entertainment,such as high resolution video,videophone,video conference,and so on.Video coding technology,which is video compressing based on,has achieved considerable development in the past decades.From the earlier H.261 to current HEVC,video coding technology has established the hybrid coding framework through many generations’ evolution.In all of those coding technologies,rate control always plays a key role.In real video application,due to the limitation of band width,how to keep the coding quality under limit bitrate,and realize output stream bitrate stability,which is called rate control problem.Except for accurate rate control,how to solve the problems that video coding quality usually has a low performance under rate control mode,and rate control gives a poor support to parallel coding framework,which have become the research focuses.About how to improve the coding quality under rate control mode,this paper proposed a temporal dependent bit allocation scheme for HEVC rate control.In traditional rate control algorithm,bits are allocated to different unit according to their context complexity.In our temporal dependent bit allocation scheme,the temporal correlation is also adopted in.Specifically,if someone unit is referenced many times in the future frames,which means it has a higher temporal weight,we should consider allocating more bits to it in order to improve its coding quality.This way can ensure that all the temporal related units’ coding quality.This paper describes the algorithm procedure and implementation issues of this temporal dependent bit allocation scheme in detail.Experiments validate the effectiveness and analyze the coding complexity.About how to improve the support to parallel coding under rate control mode,this paper proposed a joint rate control scheme for multi-machine parallel coding framework.Through independent frame bit allocation and adaptive slice bit allocation,this scheme can achieve enough rate control accuracy under parallel coding.As we know,the fashion of parallel coding framework brings the challenge of data dependency,and higher parallel degree will lead to higher data correlation unavailable.This brings a huge bad influence to accurate rate control,because rate control exactly bases on the temporal and spatial correlation of video data,which makes it possible that dynamically adjusting the coding parameter according to history coding information to achieve global rate stabilization.The joint rate control scheme proposed in this paper through periodically collecting independent coders’ information and reset appropriate coding parameter after synthesize all coders’ real-time coding state.Experiments validate that this scheme can effectively keep output rate under stable target bitrate.更多还原