摘要
针对传统的三维视频稳定算法计算量较大的问题,提出了一种基于感兴趣区域内容保护的变形算法。首先,估计每个输入帧特征点的三维信息并确定感兴趣区域;然后,利用基于ROI整帧的图像变形算法和保持显著性的图像变形算法去除输入帧的抖动;最后,利用交叉方法去除ROI边界伪迹。在5个HD视频序列上的仿真实验结果表明,该算法可获得与其他先进算法相同的去抖动性能,且算法执行时间降低了至少14%。
For the issue that it exists large amount of calculation in traditional 3 D video stabilization algorithms,a warping algorithm based on content protection in region-of-interest is proposed.Firstly,the 3D information of each input frame feature points is estimated so as to determine ROI.Then,warping algorithm based on ROI whole frame and keeping the significant is used to remove the jitter of input frame.Finally,intersect method is used to remove the boundary artifact of ROI.Simulation experimental results on five HD video sequences show that proposed algorithm can get same stabilization performance with state-of-the-art algorithm,and it has reduced above 14% executing time.
出处
《电视技术》
北大核心
2015年第11期41-45,150,共6页
Video Engineering
基金
河南省科学技术计划发展项目(122400450245)
关键词
三维视频
数字图像稳定
感兴趣区域
变形算法
特征点
three dimensional video
digital image stabilization
region-of-interest
warping algorithm
feature point