摘要
光流法是一种广泛应用的像素级非刚性图像配准方法。该方法对发生平移、旋转以及放缩变换的图像能够很好地配准,但是光流法在待配准图像比较大且初始位置远离目标位置时,配准速度及位置较差。文章提出了结合模板匹配和光流法的改进算法,首先通过模板匹配算法进行粗定位,再使用光流法进行精确定位。实验表明,该方法大大提高了原算法的配准能力。
Optical flow method is a pixel-level,non-rigid method which is widely used in image registration.This image registration algorithm is robust to affine transform,such as translation,rotation and scale translation,this algorithm is sensitive to the initial location of the template image.This is a local optimal algorithm.But,when the matching image is large and the initial location of the template image is far from the target location,the speed of fitting process will be slow and the fitting result will be bad.In this paper,an improved algorithm based on template matching and optical flow is proposed.First,we estimate the location of the template image roughly,then we use optical flow algorithm to locate the target location accurately.Experimental results show that the method greatly improved the ability of the original registration algorithm.
出处
《无锡职业技术学院学报》
2010年第6期43-45,共3页
Journal of Wuxi Institute of Technology
关键词
光流法
LK算法
模板匹配
图像配准
optical flow
LK algorithm
template matching
image registration