HEVC熵编码技术研究及并行算法设计

发布时间：2018-07-12 12:38

本文选题：高效视频编码 + 熵编码　；参考：《北京邮电大学》2014年硕士论文

【摘要】：高效视频编码(HEVC)是H.264/AVC的下一代视频编码标准。该标准由ITU-T的视频编码专家组(VCEG)和ISO/IEC的运动图像编码专家组(MPEG)组成的视频编码联合组(JCT-VC)联合开发。其核心目标是在H.264/AVC的基础上,将视频的压缩效率再提升一倍,即用相同的码率实现更高质量的视频压缩。 HEVC中使用的熵编码技术是基于上下文的自适应二进制算术编码(CABAC),这种技术与H.264/AVC中的CABAC编码技术从本质上来说是一脉相承的。不同于H.264/AVC中分别应用与基本档和高档编码配置的两种熵编码算法,HEVC仅采用这一种熵编码模式。 HEVC以算法的复杂度的上升来换取压缩效率的提高,目前难以满足实际应用中实时编码的需求。从总体编码框架来看,熵编码模块由于数据间的强相关性成为了整个编码框架并行加速的瓶颈。本文基于对HEVC熵编码关键技术和编码流程的分析,提出了相应的熵编码优化算法和熵编码并行处理策略。本文首先对HEVC熵编码的编码流程进行了全面的分析。针对熵编码模块中存在的计算方式有待优化、未能考虑人眼主观特性的问题,本文给出了一种二进制算术编码器的优化算法和一种基于熵编码预估计的感知率失真优化算法。试验结果表明算术编码器优化算法对熵编码器有3%到12%的加速,而压缩效率损失在0.02%以内；率失真优化算法得到的编码图像相比于HEVC官方参考代码在主观质量上有较明显的提高。同时,考虑到HEVC熵编码模块高度串行化的特点,本文从三个层面对并行处理的可行性进行了分析。在此基础上,本文利用多线程计算技术实现了一种HEVC熵编码并行计算框架。实验结果表明,该并行算法框架相对于传统的串行算法框架具有65%到70%加速效果,同时在编码图像的主客观质量上并没有太大的损失。
[Abstract]:Efficient video coding (HEVC) is the next generation video coding standard of H.264 / AVC. This standard is jointly developed by ITU-T Video coding expert Group (VCEG) and ISO / IEC Motion Image coding expert Group (MPEG), a joint video coding group (JCT-VC). Its core goal is to double the efficiency of video compression based on H.264 / AVC. The entropy coding technique used in HEVC is context-based adaptive binary arithmetic coding (CABAC), which is essentially the same as the CABAC coding technique in H.264 / AVC. Different from the two entropy coding algorithms used in H.264 / AVC, such as basic coding and high-grade coding configuration, HEVC only uses this entropy coding mode. HEVC gains compression efficiency by increasing the complexity of the algorithm. At present, it is difficult to meet the demand of real-time coding in practical applications. From the point of view of the overall coding framework, the entropy coding module has become the bottleneck of parallel acceleration of the whole coding framework because of the strong correlation between the data. Based on the analysis of the key techniques and coding flow of HEVC entropy coding, the corresponding entropy coding optimization algorithm and entropy coding parallel processing strategy are proposed in this paper. Firstly, the coding flow of HEVC entropy coding is analyzed. Aiming at the problem that the calculation method in entropy coding module needs to be optimized and the subjective characteristics of human eyes are not considered, a binary arithmetic encoder optimization algorithm and a perceptual rate-distortion optimization algorithm based on entropy coding preestimation are presented in this paper. The experimental results show that the arithmetic encoder optimization algorithm accelerates the entropy encoder by 3% to 12%, and the compression efficiency is less than 0.02%. Compared with the official reference code of HEVC, the coding image obtained by the rate-distortion optimization algorithm has a better subjective quality than that of the HEVC official reference code. At the same time, considering the characteristics of high serialization of HEVC entropy coding module, this paper analyzes the feasibility of parallel processing from three aspects. On this basis, a parallel computing framework of HEVC entropy coding is implemented by using multithread computing technology. The experimental results show that the parallel algorithm framework has 65% to 70% acceleration effect compared with the traditional serial algorithm framework, and there is no great loss in the subjective and objective quality of the coded image.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】