视频编码中容错率失真优化技术研究
时间:2018-07-09 09:11 来源:毕业论文 作者:毕业论文 点击:次
| 摘要因特网和无线网络等传输信道的不可靠性对视频编码的高效性和传输速度提出了严峻的挑战，因为在有噪信道中传输压缩视频流容易产生严重的错误累积，从而影响视觉效果。目前绝大多数网络本身提供了差错控制机制，但应用于视频传输时普遍存在过保护或欠保护的问题，有时甚至需要反馈信道的支持。25456 本文主要研究视频编码中的容错率失真优化技术，本文的主要研究成果包括以下几个部分： (1) 提出了高效低复杂度的端到端失真估计算法。端到端失真的估计是实现容错编码的关键，传统的方法通常假定端到端失真为信源失真和信道失真之和，然而在信道错误率较大时，这一假设并不准确。本文基于理论推导，将端到端失真划分为信源失真、错误积累失真和错误隐藏失真，并进一步划分为若干个失真项，每个失真项都具有明确的物理意义，可以直接或递归计算。所推导的模型可以跟踪来自先前帧中的错误积累，即使信道错误率较大仍然保证估计的准确性，并且可以消除子像素运动补偿、环内滤波等技术对理想模型的影响。 (2) 提出了联合信源-信道率失真优化及编码控制策略。本文通过扩展传统率失真函数的定义，在率失真优化中引入误码因素，根据当前宏块的量化系数、运动矢量、可恢复性和编码所用的比特数确定当前编码模式下的编码代价，在同时考虑编码效率和容错性能的前提下实现总体优化编码。基于提出的端到端失真模型，首次精确推导出端到端失真与码率之间的关系，并得到丢包环境下的拉格朗日乘子。基于端到端失真模型与导出的拉格朗日乘子，提出联合信源-信道率失真优化算法的一般框架。 (3) 提出了基于率失真优化的混合容错编码方法。在视频编码中，多种不同的容错工具可以分别实现某种程度的差错控制，但其联合实现往往达不到预期的目的。本文通过对常用容错工具的性能分析，提出不同容错编码工具的联合优化算法，其核心在于将提出的端到端失真模型扩展到混合容错编码框架中，采用基于大数定理的方法估计与错误隐藏有关的失真项。针对灵活宏块排序和数据划分两种容错工具，进一步提出结合率失真优化的混合容错编码方案。 关键词 视频压缩，容错编码，视频传输，率失真优化，端到端失真估计 毕业设计说明书外文摘要 Title Error Resilient Video Coding Techniques Abstract The transmission of compressed video over wireless networks as well as the Internet presents many new challenges due to the problems caused by channel errors. Most of today’s video coding standards are based on a hybrid coding framework. In the error-prone environment, transmitting the hybrid-coded video may suffer from the error propagation and thereby lead to error drifting. While the transmission errors can be controlled in the network layer, the error-resilient source coding can also reduce the network burden and becomes even more efficient, and hence achieve better Quality of Services (QoS). This dissertation targets at the rate-distortion optimized error-resilient video coding. The typical video transmission system is first studied, including the theoretical analysis and modeling of the transmission errors. The end-to-end distortion function is then derived in terms of the source and channel characteristics as well as the employed error concealment method.Based on the end-to-end distortion model, the joint source-channel rate-distortion optimized error-resilient coding is proposed, including the selection of coding control parameters. The contributions of this dissertation are summarized as below. Firstly, a low-cost yet efficient end-to-end distortion estimation approach is proposed. The end-to-end distortion estimation plays a key role in error-resilient coding. The traditional end-to-end distortion models usually simply add the source and channel distortions, which are inaccurate especially when the error rate is large. Distinctively, the proposed distortion model associates all distortion items (including source, error-propagation and error-concealment distortions) with the error rate in theory. Each distortion item is further separated into several small distortion items that can be calculated either directly or recursively. In particular, the recursive calculation scheme is employed to trace the error propagation from all previous frames. Since the overall distortion in the proposed model is taken as the sum of several distortion items, it can suppress the approximation errors from sub-pixel motion-compensated prediction and in-loop filter. The proposed model can be easily extended to the block-level implementation. Experimental results show that the proposed end-to-end distortion model has very high precision with either block-level or pixel-level implementation, while its computational cost is also very low. (责任编辑：qin) |