【作者】 封颖；

【导师】 吴成柯；

【作者基本信息】 西安电子科技大学 ， 信息与通信工程， 2009， 博士

【摘要】 传统的视频编码标准如MPEG-X和H.26X系列,主要依靠编码器利用编码信号的统计特性来实现压缩编码,从而导致编码器的运算复杂度是解码器的5至10倍以上。近年来,一些新的视频应用如无线视频传感器网络、移动视频电话和无线视频监控等在快速地融入和改变着人们的生活。然而,这些新出现的多媒体应用对视频的编解码系统提出了与以往不同的需求,即编码器因资源有限需要尽可能地简单,而解码器拥有较多资源可进行复杂地解码运算,这对视频编、解码系统的结构设计提出了新的挑战。分布式视频编码具有与传统编码标准相反的系统结构,与上述应用需求完全吻合。20世纪70年代,Slepian和Wolf提出了分布式信源无损编码理论,随后Wyner和Ziv提出了有损分布式信源编码,这两个理论奠定了分布式视频编码的理论基础。与传统的视频编码系统结构有很大的不同,分布式视频编码系统将主要的高运算复杂度模块从编码端转移到了解码端,由解码器利用信源的相关性实现高效的压缩编码。分布式视频编码已成为国内外的研究热点,论文主要分析了分布式视频编码的技术特点,对分布式视频编码存在的问题进行了深入地研究并提出相应的解决方案。论文的主要工作及取得的研究成果如下:1.基于H.264/AVC的帧内编码模式选择、整数变换和量化的特点,提出了一种帧内编码全零块判决算法。该算法根据相邻块的编码信息,使用动态判决门限判决量化后的全零块,避免了变换、量化等复杂过程。与已有算法相比,论文算法的全零块判决率最大可以达到87.8%,提高了编码速度,满足了实时视频通信的要求。2.提出了一种改进的边信息插值算法。该算法在关键帧之间的前向运动估计时,设置代价函数阈值来保证目标运动轨迹的线性连续。对于运动平稳区域,采用双向运动补偿插值生成边信息;对于其它区域,在关键帧中确定多个候选块,并选择与待插值块的4个相邻块具有最小边界匹配误差的候选块作为边信息。最后,论文算法利用马尔可夫随机场结合最大后验概率估计对边界不连续的插值块进行空域平滑。实验表明,该算法降低了编码码率,并使得解码图像PSNR最高提高0.28dB。3.提出了一种解码端Skip编码模式判决算法。论文算法先在解码端进Skip模式判决,然后将判决信息通过反馈信道传输给编码器。在Wyner-Ziv帧编码时,Skip模式编码块无需编码,在解码端重构时只需要使用边信息填充即可;对于非Skip模式的编码块,论文算法将其按运动强度的大小分为两个集合进行单独编码。实验表明,论文算法在保持PSNR基本不变的条件下,使编码图像的码率最多降低54.29%。4.提出了一种双向分布式视频编码系统的码率估计算法。该算法位于解码端,算法利用边信息和差值分布的概率模型以及高位比特对编码比特平面的码率进行估计,并将估计码率通过反馈信道传送到编码器。实验结果表明,论文算法可以有效地提高码率估计的精度,在相同码率条件下,该算法重构图像的PSNR比原算法可以提高了0.2-0.8dB。针对上述提出的算法,论文都通过大量的软件仿真、测试及与传统算法的比较来验证其有效性和先进性。更多还原

【Abstract】 The traditional video coding standards, such as MPEG-x or the H.26x recommendations, rely on the encoder to exploit the statistics of the source signal. So the encoder is typically 5 to 10 times more complex than the decoder. In recent years, the emerging applications, such as wireless sensor network, mobile camera phones, and wireless video surveillance systems, have a significant impact on our lives. It breeds challenge to the coding structure characterized by a significantly lower complex encoder and a higher complex video decoder. However, distributed video coding is ideally meeting this requirement.The information-theoretic bases of distributed video coding were established in the 1970s by Slepian and Wolf for distributed lossless coding and by Wyner and Ziv for lossy coding. Compared to the traditional codec, a dramatic change is shifting the majority of complexity to the decoder. The major task of exploiting the source redundancy is partially or wholly placed at the decoder.The main contributions and innovation points of the thesis are as follows:1.On analysis of the characters of intra mode selection, integer transform and quantization, an efficient algorithm is proposed to detect all-zero blocks of the encoding intra frame in H.264. In the proposed method, dynamic selection of the threshold is used to examine the all-zero blocks in order to avoid the integer transform and quantization. Compared with the previous methods, the maximum percent of the all zero-blocks detected by the proposed algorithm in the intra macroblocks is 87.8%. It provides significant improvement of coding speed and can meet the needs of the video communication in real time.2.An improved algorithm is proposed for the side information generation process. For the blocks with low motion, the side information is interpolated by bi-directional motion compensation. For other blocks, it is replaced by a candidate block from the key frame which has minimum border error with four adjacent blocks. And the Markov random field (MRF)-Maximum a Posteriori (MAP) is used to smooth the blocks with high border error to the adjacent blocks. Simulations results show the proposed scheme can achieve 0.28 dB gains on PSNR with declining bitrate.3.A skip mode is introduced in the Wyner-Ziv frame coding process, as well as a theoretical analysis on the benefits of the block partition to the coding efficiency. The skip mode decision is introduced at the decoder on a block-by-block basis. In the proposed scheme, a mode decision process is performed to determine whether to apply skip mode to the blocks with low motion first. With the skip-mode, the block is reconstructed from the side information in the decoder. In addition, the non-skip mode blocks are divided into two parts by motion activity. The encoding bitplane is extracted within each set and encoded independently. Simulations show the proposed scheme can achieve up to 54.29% bitrate savings without visible PSNR sacrifice.4.With a comparison on a few rate estimation algorithms, a proposed algorithm for the bi-directional distributed video coding codec is proposed in the thesis. The proposed algorithm estimates the rate of each bitplane with the side information and the Laplacian distribution of difference between the source and the side information. And the result is transmitted to the encoder via a feedback channel. The simulation result indicates that the algorithm shows an efficient improvement in the precision of rate. And the complexity of the decoder is further declined.All of the proposed algorithms in this thesis are simulated and examined, then compared with conventional schemes in order to prove their validities and advantages.更多还原