多种运动补偿算法在电子稳像中的比较及应用

Comparison and Application of Motion Compensation Algorithms in Video Stabilization

对Kalman滤波算法、均值滤波算法、高斯算法、曲线拟合算法、阻尼系数算法进行了比较和分析。采用Laplacian位面法得到多个抖动视频的运动估计向量作为运动补偿对象,对上述算法进行验证,为电子稳像应用中运动补偿算法的选取提供了参考依据。结果表明,阻尼系数算法除了系数不能自动调整以外,补偿效果较好。最后针对阻尼系数算法的不足,提出了一种改进的自适应阻尼系数运动补偿算法,该算法相对于改进前具有更强的稳健性和更好的补偿效果。

Motion compensation algorithms are compared and analyzed, including Kalman filter algorithm, average filter algorithm, Gaussian kernel filter algorithm, parabolic fitting algorithm, and damping coefficient algorithm. Motion estimation vectors of several fluctuated video are obtained by Laplacian bit-plane method, which are utilized to test the above compensation algorithms. The conclusion for the preceding five algorithms conduces to select appropriate motion compensation algorithms in video stabilization. It is verified that damping coefficient algorithm is a quite effective algorithm, except that its damping coefficient can not be adjusted automatically. An improved adaptive damping coefficient algorithm is presented, which is more robust and effective.