自然科学版
陕西师范大学学报(自然科学版)
数学与计算机科学
全景图像镶嵌中运动目标的重影处理方法
PDF下载 ()
叶偲, 李良福*, 焦婷
(陕西师范大学 计算机科学学院, 陕西 西安 710119)
叶偲,女,硕士研究生,主要研究方向为图像镶嵌.E-mail:ys_stoh@126.com.
摘要:
为解决传统的图像镶嵌方法在出现局部运动目标的情况下产生的重影现象,提出一种改进的针对运动目标重影去除的方法.对于N帧图像,算法首先通过SIFT特征点聚类和运动特征点剔除方法过滤这些图像特征点,在消除参数累积误差之后,通过三帧差法分离出背景并使用无缝融合方法完成前N-1帧图像的背景融合;然后把最后一帧与前N-1帧的已镶嵌图像进行匹配和融合,得到最后的全景镶嵌图像.实验结果表明:该算法能有效解决全景图像镶嵌中运动目标的重影问题,并保留当前最后一帧图像的运动信息.
关键词:
运动目标提取; 特征点聚类; 图像镶嵌; 重影去除
收稿日期:
2014-01-02
中图分类号:
TN911.73
文献标识码:
A
文章编号:
1672-4291(2014)04-0016-06
基金项目:
国家自然科学基金资助项目(61201434); 教育部高等学校博士学科点专项科研基金项目 (20090202120002); 中国博士后科学基金特别资助项目(200902593).
Doi:
A ghost removing method for panoramic image mosaic with moving objects
YE Si, LI Liangfu*, JIAO Ting
(College of Computer Science, Shaanxi Normal University, Xi′an 710119, Shaanxi, China)
Abstract:
In order to solve the problem of significant ghost phenomenon produced in image mosaic results caused by moving objects occur in image sequence, an improved method which is mainly used to eliminating of ghosting is proposed.With this method, after SIFT feature points clustering algorithm is employed, the geometric similarity relation between matching pairs is used to eliminate feature points on moving object, thus accurate feature point matching pairs filtered out. After eliminating cumulative errors, use three frames difference method to extract the background of the previous N-1 frames and fuse them together, then match the last frame with it to produce the final panoramic mosaic image. The result shows that the improved method can efficiently remove the ghost occur in the panoramic images and keep the moving information of the last frame.
KeyWords:
moving object extraction; feature points clustering; image mosaic; ghost removing