论文部分内容阅读
在分布式视频编码(DVC)中,边信息的质量对视频序列的压缩效果有很大影响。本文中,一种基于分层运动估计的边信息提取算法引入到DVC系统中。首先,采用了两种运动估计模式,向前运动估计和双向运动估计,然后利用分层变块大小运动估计算法获得了比较准确的运动矢量。接着运动矢量滤波器修正了由于局部相似造成的错误运动矢量。最后,选择决策模块更好地处理了帧内插技术中的多种估计模式的问题。这些算法为DVC提供了更好的边信息,提高了DVC系统的性能。实验结果显示,基于这种新算法的DVC系统高于传统视频编码帧内模式 3-4dB,而仅仅低于传统视频编码IBIB模式0.5-1dB。
In distributed video coding (DVC), the quality of the side information greatly affects the compression of the video sequence. In this paper, an edge information extraction algorithm based on hierarchical motion estimation is introduced into the DVC system. Firstly, two motion estimation modes, forward motion estimation and bidirectional motion estimation are adopted, and then a more accurate motion vector is obtained by using the hierarchical motion estimation algorithm. Motion vector filters are then corrected for erroneous motion vectors due to local similarity. Finally, the decision-making module better addresses the problem of multiple estimation modes in frame interpolation techniques. These algorithms provide the DVC with better edge information and improve the performance of the DVC system. The experimental results show that the DVC system based on this new algorithm is 3-4dB higher than the traditional video coding intra mode, but only 0.5-1dB lower than the traditional video coding IBIB mode.