首页 | 本学科首页   官方微博 | 高级检索  
     检索      

全局运动估计的阈值可变双迭代法
引用本文:陈韩锋,戚飞虎.全局运动估计的阈值可变双迭代法[J].上海交通大学学报,2004,38(1):1-4.
作者姓名:陈韩锋  戚飞虎
作者单位:上海交通大学,计算机科学与工程系,上海,200030
基金项目:国家自然科学基金资助项目(60072029)
摘    要:提出了一种全局运动区域检测与全局运动模型参数估计的阈值可变双迭代法,该方法包含两个迭代过程:①迭代最小二乘法进行参数的估计,迭代过程中用可变的百分比阈值去除局部运动点的干扰:②检测完整的全局运动区域并得到最后的模型参数.阈值可变双迭代法对不同视频的适应能力较强,得到的全局运动区域更完整,模型参数的估计更稳定,实验结果验证了该方法的可行性和优越性.

关 键 词:全局运动估计  摄像机运动模型  最小二乘法  可变阈值
文章编号:1006-2467(2004)01-0001-04
修稿时间:2002年12月23

A Double-Iteration Algorithm with Variable Threshold for Global Motion Estimation
CHEN Han-feng,QI Fei-hu.A Double-Iteration Algorithm with Variable Threshold for Global Motion Estimation[J].Journal of Shanghai Jiaotong University,2004,38(1):1-4.
Authors:CHEN Han-feng  QI Fei-hu
Abstract:A double-iteration algorithm with variable threshold was proposed to detect global motion regions (GMR) and estimate parameters of the global motion model (PGMM). The proposed algorithm comprises two iterative processes. Firstly, the PGMM is estimated with an iterative least squares algorithm. A variable threshold is introduced to eliminate the disturbance of local motion. Then, the GMR is detected and the final PGMM is calculated on the GMR in the second iterative process. The proposed algorithm is adaptive to various sequences and accurate in getting the GMR and PGMM.The experiments show the validity and advances of the algorithm.
Keywords:global motion estimation  camera motion model  least squares algorithm  variable threshold
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号