【作者】 胡庆；

【导师】 陈健；

【作者基本信息】 西安电子科技大学 ， 通信与信息系统， 2015， 硕士

【摘要】 传统的视频压缩技术,以ISO/IEC MPEG-x和ITU-T H.26x为主,因其较高的压缩效率得到普遍的认可和广泛的应用。然而,上述两种视频压缩标准都是采用非对称的结构,其编码端的复杂度是解码端的5-10倍。随着无线多媒体技术的发展,这种不对称的编解码结构不能有效地满足一些低时延、低能耗要求的新兴场景,如无线视频传感网络、能量有限的视频监控场景等。分布式视频编码(Distributed video coding,DVC)方案弥补了传统方案的上述不足。在Slepian-Wolf和Wyner-Ziv两大理论的基础上,DVC通过独立编码和联合解码成功地将大量的复杂计算从编码端转移到了解码端,从而更好地适应编码端资源受限的场景。根据有无反馈信道的存在,现有的DVC方案可分为两种:解码端码率控制方案和编码端码率估计方案。前者通过反馈信道反复向编码端索取校验信息直到成功解码,虽然性能好,却有较大的传输时延。后者克服了时延问题,但是按位平面的码率估计方案不仅要求较高的估计精度,还需要在编码端准确估计出与解码端高相似度的边信息(SI)。针对这两点,本文提出了低复杂度的无反馈DVC方案来提高视频压缩性能。主要的创新点有三个部分:1)针对多编码模式中各个模式判决的准确性与其计算复杂度的矛盾,本文综合利用时间和空间相关性,提出自适应编码模式选择算法。从而降低码率和生成SI的复杂度。2)针对基于位平面和基于块的码率估计具有高精度要求的特点,本文提出了位平面重组方案,使每个位平面的码率均匀化,降低码率估计的难度。同时位平面重组之后,解决了由于多编码模式造成的LDPCA编码器编码单元长度不一致的问题。3)针对视频序列运动强烈程度不同,单纯的基于拉普拉斯的相关噪声估计模型不能准确估计出所有情况的特点,本文通过对编解码两端边信息和量化矩阵进行统计分析,并在此基础上提出基于查找表的帧级别码率估计方案,进一步增强码率估计准确度。实验表明,相比于本文参考的无反馈DVC所提方案能有效提高系统的率失真性能,同时保证编码端较低的复杂度。更多还原

【Abstract】 The traditional video coding solutions, represented by the ISO/IEC MPEG-x and ITU-T H.26 x standards, with high compression efficiency have been widespread concerned. However, they rely on a highly complex encoder to exploit the temporal and spatial redundancy. This may be not optimal for some practical scenarios with limited transmission delay and energy, such as wireless video sensor networks and low-power surveillance.Hence, the distributed video coding(DVC), based on Slepian-Wolf and Wyner-Ziv theorems, was proposed to shift the computation and storage burdens to the decoder. It encodes the video frames independently and decodes them jointly, which turns out to be an optimal scheme in scarce resource scenarios.Available DVC codecs are mostly based on decoder rate control(DRC) scheme where the parity bits for decoding can be achieved over a feedback channel. However, the frequent requests through feedback cause transmission delay. The feedback-free DVC, relying on encoder rate control(ERC) in literatures, has overcome these shortcomings. But it is usually carried out on the coding unit of bit-plane, which requires high precision of bit rate estimation and high quality of side information(SI) approximation at the encoder. In this paper, we propose to improve the coding efficiency under low encoder complexity. The innovation includes three parts: 1) an adaptive coding model selection(ACMS) algorithm, utilizing both temporal and spatial correlation, is proposed to lower the parity bits rate and the complexity in generating SI frame; 2) a bit-plane rearrangement method is used to make the coding rate on each bit-plane homogeneous. As a result, rate estimation needs to be performed only once for all bit-planes within one frame; 3) a frame-level parity bits estimation scheme is presented to enhance the efficiency of rate estimation on the basis of a lookup table. Numerical results indicate a remarkable improvement of rate distortion(RD) performance relative to the ERC framework our codec based on under low-complexity.更多还原