面向网络PVR操作的MPEG-2视频码流转码研究.pdf

算过程中，通过合理安排预测运算步骤充分利用宏块间共用信息以提高 运算效率。 在转码实现的重编码过程中，运动估计的运算量是最大的，常常占 用整个编码运算复杂度的60%～70% 。利用合理的算法通过这些原始码 流解析得到的运动预测信息重建出目标码流需要的运动矢量，从而避免 计算复杂度非常高的全局搜索运动矢量估计，提高转码效率。本文介绍 了双线性插值算法、FDVS 算法和ADVS 算法。考虑邻域内宏块的运动 具有一致性的统计特性，本文在FDVS 算法上进行了改进，提出了基于 宏块邻域运动一致性算法。对于宏块的编码方式本文也提出了快速选择 的算法，该算法以与重叠面积最大的主导宏块编码方式一致为基本原 则，如果四个宏块中至少有一个采用帧内编码方式，则目标宏块优先采 用帧内编码方式。在对图象质量要求较高的应用中，可以在重估计合成 运动矢量基础上，在较小区域内搜索使运动矢量精细化，本文提出了变 长十字搜索的精细化算法，并通过实验与现有的 HAVS 和VSS 算法进 行了比较。 关键词：MPEG ，转码，跳帧，DCT-MC ，运动矢量重估计 THE STUDY ON MPEG-2 VIDEO STREAM TRANSCODING FOR NETWORK PVR MANIPULATION ABSTRACT With the blooming development of the Internet, network has gradually become an essential exchange platform. In order to make use of this transmission platform to a full extent, the aim of video coding standard has gradually changed: From the storage oriented MPEG-1 to both storage and transmission oriented MPEG-2, to the transmission oriented MPEG-4. The new standard H.264/AVC has made the abstract layer of the network a part of the standard to meet the requirement of transmitting media stream through the internet. With the tendancy of popularization of the Internet, the construction of high performance broadband network accelerates the development of HDTV. In the practical IPTV, the transport media stream transmitting through broadcasting network is coded according to the MPEG-2 standard which the Digital TV is coded in order to maintain the quality of the images. In the application of Time-shifted TV, the original stream is transcoded while the frames in the stream are dropped alternately to meet the requirement of the interactive PVR manipulation over the network. The framework and the flow of transcoding for PVR a