【作者】 杨亚东；

【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】 近年来，基于小波变换的图像编码技术取得了长足的进步。优秀的小波图像编码器不仅具有较高的压缩效率，而且产生的码流具有可伸缩性，从而使得渐进传输和动态带宽适应成为可能。但这种码流对包丢失和信道误码很敏感，传输过程中的任一比特错误都可能导致位于其后的码流数据无法正确解码，造成恢复图像质量的大幅降低或完全无法正确解码。现有的Internet和无线网络是有噪信道，不能实时提供可靠的数据传输。因此，研究噪声信道中的图像可伸缩码流实时传输问题就成为一项非常具有挑战性的热点研究课题。 基于前向纠错的差错控制策略是解决实时通信中图像和视频可靠传输问题的重要手段。本文重点研究包删除信道和无线信道中图像可伸缩码流的抗差错保护及其高效信源信道码率分配算法。主要的研究成果和创新点如下： 1．提出了一种基于双向局部搜索的快速码率分配算法。该算法适用于基于交织器结构的抗丢包不等保护(ULP)系统，通过合理地选择搜索起始点，采用自适应邻域的方法交替进行前向和后向局部搜索，从而在获得同等保护性能的同时，减小了计算复杂度。 2．提出了一种基于估计的快速最优交织器参数选择算法。该算法适用于给定传输码率的ULP系统，通过易求得的失真优化解的期望失真的上下界估计最小期望失真对应的交织器参数，避开了复杂度较高的失真优化解的多次求解步骤。实验结果显示，此算法兼有复杂度低和鲁棒性好的特点。 3．提出了有效期望质量的概念，定义为扣除传输失败情况后接收端得到的期望质量。利用有效期望质量与等保护和不等保护相结合的策略，提出了两种ULP系统的码率分配算法：(1)基于有效期望质量最大化判据的算法。该算法以端到端有效期望质量最大作为码率分配的目标；(2)提供QoS保障的码率分配算法。该算法在传输失败概率小于给定概率阈值的约束条件下，以端到端有效期望质量最大作为码率分配的目标。两种算法都采用双向局部搜索的方法进行快速码率分配。有效期望质量性能判据克服了原期望质量性能判据没有考虑收端用户最小图像质量要求的缺陷，实验表明，这两种算法都能提高有效期望质量，降低传输失败概率，而后者能够保证传输失败概率小于用户给定的概率阈值，从而为图像通信系统提供服务质量保障。 4．提出了一种基于信源符号重排的ULP系统的码率分配方案，导出了考虑重排效果的端到端期望质量计算式，给出了相应的码率分配算法。实验表明，该算法在包数较小而包长度较大时，能有效提高ULP系统的期望质量。 5．提出了两种基于预测的乘积码系统快速码率分配算法。该两种算法把有效期望质量性能判据引入到无线衰落信道的乘积码保护系统中，解决了原系统码率分配算法传输失败概率较大的问题。实验表明，此两种算法兼有复杂度低和预测精度高的特点。更多还原

【Abstract】 Recent years, image coding technique based on wavelet transform has made great progress. While the higher compressed rate is acquired than ever before, advanced wavelet based image coder produces scalable bitstreams and makes the progressive transmission and adaptability to dynamic bandwidth enable. However, this bitstreams are very sensitive to packets loss and bit errors in transmission channels. Even if a single bit error occurs, it is highly probable that all remaining bits will not be interpreted correctly. On the other hand, the Internet and wireless networks nowadays are lossy and cannot provide reliable data transmission in real time. As a result, one challenging problem is how to transmit scalable bitstreams reliably and efficiently over noisy channels.The error control strategy based on forward error correction is one of the important methods to cope with the problem of reliable transmission of image and video in real time. In this dissertation, special emphasis is put on the error protection and high efficiency code-rate allocation algorithms for scalable bitstreams over packet erasure channel and wireless channel. The main contributions and innovation points are as fellows:Ⅰ. For the unequal loss protection (ULP) system based on interleaver structure, a bidirectional local search algorithm with adaptive neighborhood is proposed to allocate the bit budget between the source and channel coders. By rationally selecting the initial point for the search and alternately using forward and backward search with adaptive neighborhoods, new algorithm reduces the computing complexity under the comparative protection performance.Ⅱ. For the problem of code-rate allocation for the ULP system with given transmission rate, an algorithm based on estimation to select the optimal interleaver parameters is proposed. Instead of computation of each ULP distortion-optimal solution for each possible choice of interleaver parameters, it estimates the expected distortion of each ULP distortion- optimal solution from its upper and lower bounds, which results in the speedup of the search process for the optimal interleaver parameters. The simulation shows that the algorithm has low computing complexity and is more robust.Ⅲ. A new objective function named the effective expected quality is proposed, which is defined as the expected quality excluding the contribution from failure transmission. Combining equal and unequal loss protection strategy, two algorithms are proposed as following: 1. a code-rate allocation algorithm is presented to maximize the effective expected quality. 2. A code-rate allocation algorithm is proposed to maximize the effective expected quality under the constraint condition that the probability of failure transmission is below a given threshold. For these two algorithms, relevant fast bidirectional local search methods are developed. The new objective function can overcome the disadvantage existing in the old objective function based on expected quality where the minimum image quality requirement is not considered. Numerical experiments show that the above two algorithms can provide higher effective expected quality and lower probability of failure transmission with significantly lower computational complexity compared to the previous algorithms, and the latter can provide QoS provision for a image communication system.Ⅳ. A new code-rate allocation algorithm based on rearranging the source symbols within the runs of the same channel code rate is proposed for the ULP system. The relevant probabilistic formulas are developed to describe the expected peak signal-to-noise ratio (PSNR) at the receiver based on the rearrangement. The experiment results show that the new algorithm can provide significantly higher expected PSNR especially in the cases of small number of packets and large packet length compared to the previous algorithm.Ⅴ. Two code-rate allocation algorithms based on prediction for the system of product code error protection over wirwless fading channel are presented. By introducing the new objective function based on the effective expected quality, the two algorithms overcome the problem of potential high probability of failure transmission occurring to the existing system. The simulations show that the new algorithms can provide high prediction precision, and their running speeds are faster than the previous algorithms.更多还原