【作者】 王毅；

【导师】 门爱东；

【作者基本信息】 北京邮电大学 ， 电子与通信工程， 2014， 硕士

【摘要】 近年来,随着我国互联网和移动通信网络的迅速发展,多媒体通信技术得到了日益广泛的应用。然而信道环境不佳导致的丢包或误码会对视频的解码质量造成严重影响。因此,提升视频码流的鲁棒性的丢包补偿技术成为视频编码技术中重要的研究方向。本文基于H.264/AVC标准对其丢包补偿技术进行研究,并提出改进算法。本文结合计算机图形学和计算机视觉领域中的图像修复算法和H.264/AVC已有空间域错误隐藏算法提出了一种基于样本和快速结构化的空间域错误隐藏改进算法。经过仿真测试,与H.264/AVC已有空间域错误隐藏算法相比,本文提出的算法在客观指标上比已有算法平均提高0.93dB,主观效果比已有算法好。本文结合视频修复的关键算法运动矢量修复和H.264/AVC已有的时间域错误隐藏算法提出了一种基于帧间模式和运动矢量修复的时间域错误隐藏改进算法。经过仿真测试,与H.264/AVC已有时间域错误隐藏算法相比,本文提出的算法在客观指标上比已有算法平均提高0.58dB,主观效果比已有算法好。更多还原

【Abstract】 In recent years, with the development on the Internet and mobile communication networks, multimedia communication technology has increasingly extensive usage. However, the packet loss or bit error caused by unsteady channel environment can heavily undermine the video decoding performance. Therefore, the packet loss compensation technology, which can improve the robustness of video stream, becomes an important research direction in video coding technology. Based on the study of H.264/AVC standard and its packet loss compensation, we propose two improved algorithms.In this thesis, by combining the image inpainting algorithms in computer graphics and computer vision and the spatial error concealment algorithms in H.264/AVC standard, we first propose an improved spatial error concealment algorithm based on exemplar and fast marching image inpainting. After the simulation and test, compared with the spatial error concealment, our proposed algorithm has a0.93dB improvement in objective performance in average and also has a better performance in subjective quality.Then, by combining the motion inpainting algorithms in video inpainting and the temporal error concealment algorithms in H.264/AVC standard, we propose an improved temporal error concealment algorithm based on inter mode and motion inpainting. After the simulation and test, compared with the temporal error concealment, our proposed algorithm has a0.58dB improvement in objective performance in average and also has a better performance in subjective quality.更多还原