傅里叶变换域的视频运动稳定评价方法

Video motion stability evaluation method based on Fourier transformation domain

针对不同稳定程度的视频,在傅里叶变换域提出了一种视频稳定性评价算法,首先计算视频的帧间仿射变换矩阵,得到在变换空间表示的相机路径;然后计算相机路径垂直投影在图像平面上的平移量和旋转量,分别得到视频的平移和旋转运动轨迹;计算视频运动轨迹的傅里叶变换,获取视频运动轨迹的傅里叶变换虚部曲线;最后根据待测视频运动轨迹和稳定视频运动轨迹的虚部曲线计算出一个衡量值,作为该视频的稳定系数,并以其评价视频的运动稳定性。实验结果表明,所提算法与主观评价结果高度一致,能够准确地评价不同稳像算法的稳像性能。

A novel video stability assessment algorithm is proposed for the video with different stability on Fourier domain.First,an affine transform matrix between frames is computed for camera path.Then,shift and rotation value is computed by vertically projecting camera path to the image coordinate system for the corresponding video motion trajectories.Next,the imaginary part of Fourier transformation for each motion trajectory is obtained by computing its Fourier transformation.Finally,a measure value as stability factor is calculated according to the imaginary curve trajectory and the ideal trajectory for the stability assessment of the video.The experiment result shows that the proposed algorithm keep consistent with subjective evaluation results,and can accurately evaluate the stability of a shaky video stabilized by different stabilization algorithms.