动态场景中自适应去除外点的全局运动估计方法

Global Motion Estimation Method with Adaptive Outliers Elimination in Dynamic Scene

为在动态场景图像序列中准确地完成全局运动估计,提出一种自适应去除外点的全局运动估计方法。对尺度不变特征变换(Scale invariant feature transform,SIFT)算法提取出的特征点利用最近邻搜索算法中的BBF(Best Bin First)方法进行匹配。为提高全局运动估计的精度,提出改进的随机抽样一致(RANdom SAmple Consensus,RANSAC)算法。此算法能够自适应地去除外点,即利用特征点运动矢量的方差控制迭代次数来进行外点的去除,最终通过摄像机运动模型实现准确的运动参数估计和背景补偿。对标准图像序列Coastguard和实际拍摄的动态场景图像序列的实验表明,提出的方法能够快速地完成动态场景中的全局运动估计与补偿,具有较高的精度和适应性。

To exactly obtain global motion estimation in dynamic scene,this paper presents an adaptive global motion estimation method to eliminate outliers.The Best Bin First（BBF） method of the nearest neighbor search algorithm is used to match feature points extracted by the scale invariant feature transform（SIFT） algorithm.In order to improve the accuracy of feature matching,an improved RANdom SAmple Consensus（RANSAC） algorithm is proposed that can eliminate outliers adaptively.The iterative number is controlled by the variance of motion magnitude of feature points.Through a camera motion model,accurate results of parameter estimation and background compensation are obtained.The proposed algorithm is tested by the Coastguard standard image sequence and the practical one with dynamic scenes.The experimental results are compared with the previous method,which demonstrates that the proposed algorithm is highly accurate and adaptive and that the speed is faster.