基于改进运动矢量估计法的视频稳像  

Video stabilization with improved motion vector estimation

作　　者：吉淑娇[1,2,3] 朱明[1] 雷艳敏[3] 胡汉平[1,2] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院大学,北京100049 [3]长春大学电子信息工程学院,吉林长春130022

出　　处：《光学精密工程》2015年第5期1458-1465,共8页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61203242);吉林省科技发展计划资助项目(No.20140101084JC);吉林省教育厅资助项目(No.2015LY502L11)

摘　　要：提出用改进的运动矢量估计方法对视频序列进行运动补偿,以有效去除机载成像设备摄取视频序列的帧间抖动,实现电子稳像。介绍了BRISK算子提取特征的模型和算子描述方式,利用BRISK算子提取相邻帧的稳定特征点并进行匹配,结合参数仿射变换模型实现帧间运动矢量粗估计。为了提高运动矢量估计的精度,引入粒子滤波解决因特征点景深不同引起的运动矢量估计失准的问题,继而采用加权最小二乘法改进运动矢量估计。最后采用卡尔曼滤波从全局运动矢量里分离出运动补偿分量,对视频序列进行逐帧补偿。实验表明,视频序列稳像后的帧间变换保真度值提高了近3db,显示提出的的稳像算法可以处理复杂运动下视频的帧间抖动,且快速准确并具有一定的鲁棒性。An improved motion vector estimation algorithm was proposed to remove effectively the inter frame jitter of the video sequences recorded by airborne imaging equipment and to realize the electronic image stability.The BRISK operator and the corresponding feature extraction model were introduced.Then,the BRISK operator was employed to extract and match the stable feature points of neighbor frames.By combining with the affine transformation motion model,the rough estimation of inter frame motion vectors was implemented.To improve the accuracy of motion vector estimation,the particle filter with the weighted least-square method was employed to solve the problem of the motion vector calculation inaccuracy caused by different depths of the feature points.Finally,the Kalman filter was used to separate the jitter components from the global motion vectors for compensating the video sequences frame by frame.The experimental result indicates that the inter-frametransformation fidelity（ITF）of video sequence has improved about 3db,which verifies that the proposed stabilized algorithm removes inter-frame jitter of high frequency under complicated movement with fast speed,higher accuracy and stronger robustness.

关 键 词：电子稳像 视频稳像 特征匹配 粒子滤波 运动补偿 评价指标 

分 类 号：V557[航空宇航科学与技术—人机与环境工程] TN941.1[电子电信—信号与信息处理]