数字图像相关方法辅助的视频运动放大改进方法及其在位移测量中的应用

An improved video motion magnification method assisted by the digital image correlation method and its application in displacement measurement

视频运动放大(Video Motion Magnification,VMM)方法因其能放大视频中的微小运动,而受到工程测量领域的广泛关注。当放大微小运动的视频时,若VMM方法中的时域滤波器参数设置不恰当,放大后的视频容易产生图像伪影和噪声。为了克服传统方法中需要手动设置滤波器参数的缺点,提出一种由数字图像相关(Digital Image Correlation,DIC)方法辅助的VMM改进方法。该方法的主要思想是由DIC方法计算图像位移并由此分析出运动主频,然后将VMM方法中的时域滤波器参数设置在包含该运动主频的较窄区间范围内。这种参数设置方法可保证放大后的视频中保留图像细节且无明显伪影。室内多频运动实验以及室外单频运动实验的结果表明,本文所提出的方法不需要反复手动设置参数试错,可直接获得清晰的运动放大后视频,并可进一步实现微小运动的位移测量。

Video Motion Magnification(VMM)has attracted extensive attention in the field of engineering measurement because of its ability to magnify tiny motion in video.When the video with small motion is magnified,if the time domain filter parameters are not set properly in VMM,the magnified video is prone to produce image artifacts and noise.In order to overcome the shortcomings of traditional methods that require manual setting of filter parameters,an improved VMM method assisted by Digital Image Correlation(DIC)method is proposed.The main idea of this method is to calculate the image displacement by DIC method,analyze the motion master frequency,and then set the parameters of VMM time-domain filter in a narrow range including the motion master frequency.This parameter-setting method can ensure that the image details are preserved in the magnified video without obvious artifacts.The results of indoor multi-frequency motion experiments and outdoor single-frequency motion experiments show that the method proposed in this paper can directly obtain clear video after motion magnification without repeatedly setting parameters manually,and further realize the displacement measurement of small motion.