Small Motion Magnification Using Automated RoI Selection and Spatial Co-ordinate Approach

This paper presents a new framework to magnify small motions present in videos, which are invisible to the naked eyes. The proposed method is based on spatial coordinates. It has an ability to magnify motions of our interest in presence of large background motions as well as static background. It is assumed that the camera is static when the video is recorded. Initially, regions of interest (RoI) are detected in our method, which need to be magnified. Histogram of oriented gradients (HoG) with thresholding is used to estimate the RoIs. Subsequently, a feature point tracker is used to detect the motion of feature points with respect to time. Finally, the velocity of these feature points are amplified by a magnification factor, which results in a video with small motions amplified. The proposed method provides a newly introduced block based spiral search technique for automatically detecting RoIs. The qualitative analysis of the experimental results shows that the proposed method can produce significant motion magnification even for very small motions. The experimental results show the efficacy of the proposed motion amplification scheme.