基于背景特征点匹配的视频稳像  

Video stabilization based on background feature points

作　　者：吉淑娇[1] 雷艳敏[1] 朱明[2] 

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

出　　处：《吉林大学学报（工学版）》2016年第6期2116-2122,共7页Journal of Jilin University:Engineering and Technology Edition

基　　金：吉林省教育厅项目(2015LY502L11);吉林省科技发展计划项目(20140101084JC)

摘　　要：针对特征点在前景运动物体上会严重影响运行矢量准确性的问题,本文提出利用特征点分类的方法,通过计算KLT跟踪特征点,计算运动估计所得匹配点跟踪所得特征点之间的距离,并由MSAC算法不断更新前景特征点和背景特征点集,最后仅采用背景上的特征点进行全局运动估计,从而进行稳像。实验结果表明,本文方法稳像后视频平滑度更高,而且本算法的无定义区域明显减少,更利于人眼视觉感受。The objective of this paper is to stabilize electronic image for aerial video with dynamic background. When shooting, as the camera follows the target, the video sequence changes scenarios. This is due to the movement of the camera and the change of background, so the scene is complicated. The difficulties in stabilizing video image with dynamic background include： an accurate estimate motion vector of camera movement and background compensation. To overcome such difficulties, this paper uses KLT track operator in combination with the M-estimator Sample and Consensus （MSAC） algorithm to realize feature point classification. Then, only the feature points on the background are used for global motion estimation to perform image stabilization. Experimental results demonstrate the feasibility of the proposed method.

关 键 词：信息处理技术 电子稳像 背景特征 运动补偿 

分 类 号：TN941.1[电子电信—信号与信息处理]