基于S变换的非平稳微小运动自动放大

发布时间：2018-08-24 14:02

【摘要】：现有的欧拉视频微小运动放大方法没有考虑视频中运动信息的自动检测,在实现运动放大时需要人为选取感兴趣运动信息处理的合适参数,比如滤波器截止频率、放大倍数等。对于一般的视频,通常无法直观地确定这些参数,而是通过试错来完成。为此,基于S变换提出了一种视频序列中非平稳微小运动的自动放大方法:基于S变换自动确定带通滤波器的即时相关参数并设计了相应的动态滤波器,在此基础上实现了视频微小运动的全自动放大。首先,通过S变换分析出视频中微小运动在各个时刻的瞬时频率;然后,对视频在不同时刻呈现不同频率的情况进行动态带通滤波处理;最后,放大经过带通滤波后的感兴趣运动信息,实现微小运动放大。此外,对于抗噪性能分析,提出了一种视频区域信噪比评价方法。实验结果表明,所提方法在对实际视频进行放大处理时,能够根据运动信息频率的变化自动地获取滤波器、放大倍数等参数,无需人工参与;运动放大后可以动态地展现运动目标的放大效果。同时,准确的动态滤波器在一定程度上能够抑制噪声干扰,使得运动放大效果更加理想。
[Abstract]:The existing Euler video micro-motion amplification methods do not consider the automatic detection of motion information in the video. When realizing the motion amplification, we need to choose the appropriate parameters of motion information processing, such as filter cutoff frequency, magnification and so on. For general video, these parameters are often not visually determined, but are done by trial and error. Based on S-transform, an automatic amplification method for non-stationary micro-motion of video sequences is proposed. Based on S-transform, the real-time correlation parameters of band-pass filter are automatically determined and the corresponding dynamic filter is designed. On this basis, the automatic amplification of video micro motion is realized. Firstly, the instantaneous frequency of tiny motion in video is analyzed by S-transform. Then, the dynamic bandpass filter is used to process the video with different frequency at different time. Small motion amplification is realized by amplifying the motion information of interest after band-pass filtering. In addition, a method for evaluating the signal-to-noise ratio (SNR) of video region is proposed for the analysis of anti-noise performance. The experimental results show that the proposed method can automatically obtain the filter, magnification and other parameters according to the change of the frequency of the moving information when the actual video is amplified. The magnification effect of moving target can be dynamically displayed after motion amplification. At the same time, the accurate dynamic filter can suppress the noise to a certain extent and make the effect of motion amplification more ideal.
【作者单位】：湖北省水电工程智能视觉监测重点实验室(三峡大学);广东省粤电集团有限公司天生桥一级水电开发有限责任公司水力发电厂;
【基金】：国家自然科学基金资助项目(61272237,61402259) 湖北省自然科学基金创新群体项目(2015CFA025) 湖北省教育厅科学技术研究计划重点项目(D20151204) 水电工程智能视觉监测湖北省重点实验室开放基金资助项目(2014KLA04)~~
【分类号】：TN713;TP391.41

【相似文献】