基于Kalman滤波的稳像技术研究

Digital Image Stabilization Using Kalman Filtering

提出一种适用于视频侦察的实时电子稳像方法。采用特征点匹配法估计帧间全局运动参数,在建立参数运动模型的基础上,通过Kalman滤波确定各帧修正矢量。实验结果说明该方法可以较好地去除高频抖动,保留摄像机的主动运动,实现实时稳像。

A real-time stabilization system using Kalman filtering for video was presented. Inter-frame global motion vectors were predicted by feature matching based motion estimation. The global camera motion was defined in terms of constant velocity motion models, and Kalman filtering was employed to facilitate smooth operation. Experimental results for many video sequences show that the method can eliminate high frequency tremble, remain active motion of camera and realize real time image stabilization satisfactorily.