自动检测及放大视频中的微小运动

Automated Detection and Magnification of Small Motion in Videos

欧拉视频运动放大方法在实现微小运动放大时需要人为选取带通滤波器的相应通频带参数.通常情况下,人们无法直接确定视频中运动目标的精确运动频率,从而使得该方法对未知的视频难以操作.为此,提出一种视频微小运动的自动检测及放大方法,首先对视频中微小运动基于时间序列的运动信息进行分析,通过功率谱估计得出运动的频率等信息,然后在整幅图像空域通过聚类分析和阈值选取,确定整个视频感兴趣运动的中心频率,并据此得出带通滤波器的参数,以实现视频微小运动的自动检测及放大.实验结果表明,本文的自动检测及放大方法不仅操作简单、取得的运动放大效果好以及支持更大的放大倍数,而且抗噪性能明显优于原始的欧拉运动放大方法以及半自动运动放大方法.

It is difficult to get the band-pass filter parameters for the conventional Eulerian Video Magnification （ EVM ） for a given video with small motion to magnify. In this paper, an approach to detect and magnify the small motion with parameters extracted from the video automatically is proposed. Firstly, small motion based on motion information of the temporal sequences is analyzed in video, and frequency parameters are obtained by the estimation of the motion points＇ power spectrum. Then, the center frequency of interested motion is determined by clustering analysis and threshold selection in the whole image, which can accordingly get the band-pass filter parameters. Finally, the automated detection and magnification of small motion has been achieved. The experimental results show that the proposed automated detection and amplification method is not only simple to operate, and can achieve good motion magnification results, but also support larger magnification factor. And its anti-noise performance is significantly better than both the original Eulerian method and the semi-automated magnification methods.