分布视频编码中基于帧间相关性的自适应关键帧选取算法

Adaptive key frame selection algorithm based on the interframe correlation in distributed video coding

针对分布式视频编码(DVC)系统中固定周期关键帧选取(PKFS)方法忽视了帧间相关性的缺陷,提出了一种自适应关键帧选取(AKFS)算法。利用图像特征点检测与匹配的方法,将相邻图像的非匹配点作为帧间相关性的近似,把累积或平均非匹配点数超过阈值的帧判定为关键帧。在此基础上,提出改进的帧内插方案,以适应不同长度序列组的边信息生成;将零运动强度的关联帧合并为一帧图像参与编解码,进一步提高了系统的压缩效率。实验结果表明,对于不同运动特性的序列,本文提出的算法可以明显提升边信息帧的重建质量,使系统的率失真性能提高0.9～2.0 dB,并有效降低了编码传输码率。

An adaptive key frame selection（AKFS） algorithm is proposed to deal with the problem of periodic key frame selection（PKFS） algorithm without considering the interframe correlation of video sequence in the distributed video coding（DVC） system.The unmatched points in the adjacent frames are used as an approximation of the interframe correlation by detecting and matching interesting points of image,and the frame,whose accumulative or average number of unmatched points is above the threshold,can be judged as the key frame.An improved scheme of frame interpolation is proposed in order to generate the side information for flexible group-of-picture（GOP） lengths.The compression efficiency is further improved after the correlated frames with zero motion intensity are merged into one frame for encoding and decoding.Experimental results show that 0.9-2.0 dB enhancement in the rate-distortion performance is achieved by applying the proposed algorithm in the DVC system,and the transmission code rates are reduced in different video sequence environments.