一种基于改进运动估计的电子稳像技术  

An Electronic Image Stabilization Technology Based on Improved Motion Estimation

作　　者：冀爽[1] 方明[1] 王成[1] 韩永林[1] 

机构地区：[1]长春理工大学计算机科学技术学院,长春130022

出　　处：《长春理工大学学报（自然科学版）》2014年第4期121-124,128,共5页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省科技发展计划项目(20120333;20130101054JC)

摘　　要：在摄像机载体运动过程中会造成图像序列的不稳定,从而产生观察者的视觉疲劳甚至会丢失观察目标,电子稳像是解决以上问题的有效手段。电子稳像首先对全局运动进行估计,求出参考帧和当前帧图像的全局运动矢量。本文先取视频序列相邻两帧图像的子图像进行块匹配得出局部运动矢量,根据计算得到的运动矢量弧度值剔除错误的运动矢量,然后对剩余的运动矢量求平均值作为此区域的运动矢量,并把与此矢量弧度相近的块的中心点作为特征点,利用仿射变换模型求解出变换参数,最后经过运动补偿技术去除视频抖动。实验结果表明,该方法可以有效地抑制帧间的平移和旋转等抖动,达到了预期的效果。Unstable camera image sequence is caused in the process of the movement of the camera＇s carrier,this visual fatigue will be produced and the object will even be losed. Electronic image stabilization is effective means to solve these problems. During the process of electronic image stabilization,at first the global motion estimation is done to calculate the global motion vector between the reference frame and the current frame. In this paper,the block matching algorithm is used for the sub-images of two consecutive frames in a video sequence. The wrong vectors can be eliminated on the basis of the calculated offset radian of the motion vector,then the average of the remaining motion vectors is calculated as the motion vector in this area and the center point of the block is selected which offset radian is similar to the motion vector as a feature point to calculate the transformation parameters by using the affine transformation. The experimental result shows that this method can restrict inter-frame translation and rotation movements effectively and the desired results can be reached.

关 键 词：电子稳像 运动估计 块匹配法 运动矢量 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]