一种视频图像目标平移运动估计算法  

A TRANSLATIONAL MOTION ESTIMATION ALGORITHM FOR TARGET IN VIDEO IMAGE

作　　者：李世扬[1] 储泽楠[2] 

机构地区：[1]安阳工学院人事处,河南安阳455000 [2]安阳工学院科研处,河南安阳455000

出　　处：《计算机应用与软件》2015年第12期184-187,200,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(U1204613)

摘　　要：运动估计是数字视频处理的基本问题之一,目前广泛采用的是相位关法,由于频谱泄漏和边界效应等原因,其运动估计的准确性难以保证。为此,在讨论基本的相位相关算法的基础上,引入一种用于构造二维圆对称窗函数的核回归方法,提出基于加窗预处理的相位相关法,并将其应用于运动目标的平移运动估计。对比实验结果表明,该方法能够有效地降低频谱泄漏和边界效应对运动估计的影响,提高了相位相关法对运动目标平移估计的正确性。Motion estimation is one of the basic problems in digital video processing; however,the phase correlation method widely used currently cannot ensure the accuracy of motion estimation due to the spectrum leakage and boundary effect or other reasons. Therefore,based on discussing basic phase correlation method,we introduced a kernel regression method for the construction of two-dimensional circular symmetrical window function,and proposed the windowing preprocessing-based phase correlation method,and applied it in translational motion estimation of moving objects as well. In contrasting experiment,result shows that the proposed method can effectively reduce the influence of spectrum leakage and boundary effect on motion estimation, and improves the accuracy of phase correlation method on translational motion estimation of moving objects.

关 键 词：相位相关 核回归 窗函数 平移运动估计 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]