摘要
在分布式视频编码中,边信息的质量对系统整体的率失真性能有着至关重要的影响。在大GOP(图像组)条件下,外推算法因其具有良好的实用性而被广泛采用。为了进一步提高在大GOP条件下利用外推算法生成边信息的质量和系统的率失真性能,提出了一种基于混合更新的边信息生成算法。该算法将外推算法和内插算法混合使用,在利用外推算法生成边信息的基础上,通过运动估计、矢量滤波和运动补偿生成内插边信息,再对两者进行自适应加权生成混合边信息并更新,同时对更新后的混合边信息和解码WZ帧进行二次重建。实验结果表明:该算法在大GOP和编码端同等复杂度的情况下,能够显著提高边信息的质量以及系统的率失真(RD)性能,且性能稳定,适用于不同运动强度的视频序列。
In the distributed video coding, the quality of side information plays an important role in the rate distortion performance of the system. The extrapolation method is widely used because of its good practicability in large GOP. In order to further improve the quality of the generated side information and the rate distortion performance with extrapolation algorithm in large GOP, an algorithm of side infor- marion generation based on hybrid updating is proposed. It is a hybrid one which consists of both extrapolation and interpolation algo- rithm. On the basis of the generated side information resulted from the extrapolation algorithm,the interpolation side information has been produced with motion estimation, vector filtering and motion compensation. While hybrid side information has been generated and updated via self-adaptive weight with the two algorithms the updated hybrid side information and WZ frame are reconstructed again. Experimental results show that the hybrid algorithm can significantly improve both the quality of side information and system rate distortion perform- ance,and that performance of hybrid algorithm is stable,which is suit for video sequences with diverse exercise intensity.
出处
《计算机技术与发展》
2017年第4期55-59,共5页
Computer Technology and Development
基金
黑龙江省教育科学技术研究项目(12541707)
关键词
边信息
混合更新
外推算法
内插算法
率失真
side information
hybrid updating
extrapolation algorithm
interpolation algorithm
rate distortion
