摘要
为解决视频拼接中源视频在时间上不同步的问题,提出了一种视频帧对齐方法。利用图像矩阵间的相关系数特性,根据视频拼接中源视频的特点,运用运动量检测方法检测视频中不同时间点的运动量,在低运动量点运用相位相关法确定交叠区域大致范围,在高运动量点计算相关系数确定同一时间点上的对应帧图像。在对应帧的搜索过程中,使用变步长搜索算法提高搜索效率,减少搜索时间。实验中使用无标定的双摄像机采集多组视频数据,结果表明了该方法的准确性和有效性。
To align the source videos in the video mosaic, one kind of frame alignment method is put forward. Based on the property of the relative-coefficient between source images and the property of source videos in the video mosaic, the movement at different time points is calculated by the movement detection firstly. Then the approximate range of overlap zone is calculated by the phase correlation method at the low movement point while the matching frame images is located as the relative-coefficient between them at the high movement point. The variable step-size search algorithm is taken to improve the searching efficiency and reduce searching time cost in the matching frame search. With the representative source videos captured by two cameras without former calibration, the validity and efficiency of the method is demonstrated.
出处
《计算机工程与设计》
CSCD
北大核心
2012年第3期1037-1041,共5页
Computer Engineering and Design
关键词
视频帧对齐
部分重叠区域
相位相关法
相关系数
运动量检测
变步长搜索算法
sequence-to-sequence alignmentl part overlapped zone
phase correlation relative-coefficient movement detection variable step-size search algorithm
