【作者】 杜建超；

【导师】 吴成柯；

【作者基本信息】 西安电子科技大学 ， 信号与信息处理， 2006， 博士

【摘要】 随着通信技术和网络技术的发展,视频流业务在网络通信中占据越来越重要的地位。然而,由于目前Internet和无线网络不能提供可靠的数据传输,因此视频流在这些网络中传输时经常会发生误码和数据包丢失等传输错误。比如Internet网络共享链路的特点使各种业务相互竞争网络资源,常常出现数据拥塞而导致丢包;无线网络频带资源有限、带宽波动较大以及无线信道多径衰落、共道干扰的特点使得信号在传输过程中既受到随机(比特)错误的影响,又受到突发错误(丢包)的影响。而且,由于目前Internet、802.11无线局域网、3G通信网、PSTN、ISDN和ATM等网络之间具有异构性,即不同的网络具有不同的带宽、丢包率等性能和特点,因此视频流在网络之间传输时也容易发生传输错误。另一方面,压缩过的视频流对传输错误非常敏感,这是因为目前所有成熟的视频压缩标准都采用VLC变长编码和运动估计等时空预测技术。因此,设计鲁棒的视频编码方案和有效的差错控制策略成为目前视频传输研究的重点内容。容错编码技术是解决视频流鲁棒传输的重要手段。本文从增强数据的抗误码性能、选择保护重要数据和提高解码端错误掩盖性能等方面入手,对基于H.264/AVC码流的视频容错技术进行了深入研究,并取得了以下成果:第一、提出了一种采用空时域掩盖相结合的Intra帧错误掩盖算法。该算法结合时域运动补偿,利用帧间相关性来预测丢失的Intra帧信息,在掩盖过程中自适应选择掩盖模式。由于Intra帧并非总是发生在场景转换时刻,因此选择时空域掩盖相结合的办法克服了空域插值掩盖算法精度不高的缺点,能够得到更好的预测效果。实验结果表明该算法比H.264/AVC的非标准错误掩盖算法大大提高了掩盖效果,改善了图像质量。第二、提出了一种基于图像内容的数据分割算法。该算法根据宏块对重建图像质量的贡献程度将每一编码帧的所有Inter宏块划分为高影响度和低影响度两组,并根据网络条件为两组宏块分别添加不同的FEC信道编码,从而优先保护高影响度的一组宏块,达到改善图像质量的目的。该算法克服了H.264/AVC标准中分割方法过于僵化的缺点,可更好地适应图像内容的变化和网络条件的改变。第三、提出了一种可降低误差扩散的数据分割算法。该算法根据编码器运动估计的结果,评估每一宏块对后续帧质量产生的影响,将所有宏块划分为高、中、低三类,并对不同类宏块的运动矢量和DCT系数应用不同的信道编码策略。由于该算法能有效保护被后续帧参考次数较多的宏块,因此能降低易错环境下视频传输时误差扩散的程度,从而改善图像质量。实验结果表明该算法能良好地平滑图更多还原

【Abstract】 With the technological development of communications and networks, video streaming services are playing a more and more important role in the field of network communications. However, the Internet and wireless network nowadays can not provide reliable data transmission, and the bit error and packet loss are liable to occur at the moment of delivering video streaming over these networks. For example, the characteristic of Internet of sharing path makes different services compete for the network resources with each other, which results in data congestion and packet loss, while the characteristics of wireless network of limited bandwidth, high bandwidth fluctuation and wireless channel of multi-path fading, interference make the signal be affected by not only random bit error but also burst error (packet loss). Furthermore, the current networks, such as Internet, 802.11WLAN, 3G communication systems, PSTN, ISDN, and ATM etc., are heterogeneous, namely different networks have variable poperties and traits in bandwidth and packet loss rate. Thus, data of video streaming are also liable to encounter transmission errors when delivered accross them. On the other hand, the compressed bit streams are extremely sensitive to transmission errors beacuse almost all of the current video compression standards adopt variable length coding (VLC) and spatial or temporal prediction techniques such as motion estimation. As a result, designing robust video coding schemes and efficient error control strategies becomes an important research content in the field of video transmission.Error resilient coding technique is an important means to solve robust transmission of video streaming. In this dissertation, special emphasis is given on the error resilient technique for H.264/AVC coding bit stream. The main contributions and innovation points are as follows:1. An error concealment algorithm based on hybrid spatial-temporal concealment for intra frame is proposed. Combined with motion compensation, the new method utilizes the correlations of inter frames to predict the lost information of Intra frames, and adaptively selects the mode of concealment. Since Intra frames are not encoded only at the moment of scene change, the hybrid spatial-temporal concealment method overcomes the drawback of low predictive precision from spatial interpolation method and achieves better results. The experimental results show that the proposed method highly outperforms the H.264/AVC non-normative error concealment method in quality更多还原