多视频超分辨率时间重建技术可以消除视频中的运动模糊和运动混叠现象。进行多视频时间重建时, 重建倍数过小不能有效地减小运动模糊和运动混叠现象; 而重建倍数过大会产生严重的时间振铃现象和运动物体位置的跳变。本文对如何在重建过程中确定最优的时间重建倍数进行了研究。通常情况下, 在决定多视频时间重建倍数时只考虑输入视频的个数; 但在实际重建过程中, 除了输入视频个数的因素, 时间重建倍数还与各个输入视频在时间上能提供的冗余信息量有关。文中从多个低分辨率输入视频配准后的位置关系入手, 通过对曝光区间的分析, 提出了确定多视频时间重建倍数的计算方法。最后对提出的方法进行了理论分析与实验, 证明了所提方法的有效性。

Video-to-video temporal super-resolution technology eliminates motion blur and motion aliasing existing in a captured video sequence. However, smaller temporal reconstruction times can not effectively remove motion blur and motion aliasing, while larger times will lead to serious temporal-ringing and the wrong location of a moving object in process of temporal reconstruction. This paper explores how to select an appropriate temporal construction times in the temporal reconstruction. In general, only the number of input video sequences is considered in deciding temporal reconstruction times of video-to-video super-resolution reconstruction. In fact, temporal reconstruction times are also associated with redundant temporal information provided by each input low-temporal resolution video sequence. So, a novel method to select an appropriate temporal reconstruction times is proposed by analyzing the exposure-time and location relation of aligned low resolution input video sequences. From theoretical discussion and experimental results, the proposed method is validated to be effective and efficient.