基于Harris-LM全局运动估计的电子稳像算法研究  

Research on electronic image stabilization based on Harris-LM global motion estimation

作　　者：王兆军[1] 田裕鹏[1] 

机构地区：[1]南京航空航天大学自动化学院,江苏南京210016

出　　处：《应用科技》2010年第1期32-35,共4页Applied Science and Technology

基　　金：航空科学基金资助项目(20090152004)

摘　　要：全局运动估计用途非常广泛,但是由于全局运动估计计算量大、噪声多等因素使得它很难在实时应用中使用.为了减少全局运动估计的计算量,提高全局运动估计的速度,选取Harris角点进行计算.基于LM(Lev-enberg-Marquardt)全局运动估计的思想,提出一种去除图像画面抖动的算法.通过LM算法估计全局运动参数,从而获得参考帧和当前帧之间的全局运动矢量.然后采用Kalman滤波的方法,一方面较好地平滑运动矢量以降低视频抖动,另一方面有效跟随摄像系统的有意扫描运动.最终进行运动补偿,以达到输出稳定图像的目的.实验结果表明,该算法能较好地实现图像序列的稳定.This paper presented an anti-jitter algorithm tion. At present, the global motion estimation is used based on LM （Levenberg-Marquardt） global motion estimawidely, however, it is hard to be applied in real-time situation due to large amount of calculation and noise. In order to simplify the calculation and increase the speed of global motion estimation, the authors choose the Harris comer points for calculation. First, estimate the global motion parameters by use of LM algorithm, and then obtain the global motion vector between the reference frame and the current frame. Next, use the Kalman filter to better smooth the original motion vectors so as to alleviate unwanted video jitter and at the same time, follow the deliberate camera scan effectively. Finally, stable images output is obtained by motion compensation. The experimental results show that this algorithm can produce stabilized series of images.

关 键 词：电子稳像 全局运动估计 Levenberg—Marquardt算法 特征匹配 KALMAN滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]