基于关键特征点提取的图像快速配准方法  

Fast Image Registration Based on Extracting Key Feature Points

作　　者：高阳 史再峰 庞科 曹清洁 姚素英[1] Gao Yang;Shi Zaifeng;Pang Ke;Cao Qingjie;Yao Suying(School of Microelectronics,Tianjin University,Tianjin 300072,China;School of Software and Communication,Tianjin Sino-German University of Applied Sciences,Tianjin 300350,China;School of Mathematical Sciences,Tianjin Normal University,Tianjin 300387,China)

机构地区：[1]天津大学微电子学院,天津300072 [2]天津中德应用技术大学软件与通信学院,天津300350 [3]天津师范大学数学科学学院,天津300387

出　　处：《南开大学学报（自然科学版）》2020年第2期56-61,共6页Acta Scientiarum Naturalium Universitatis Nankaiensis

基　　金：国家自然科学基金(61674115);国际科技合作项目(14RCGFGX00845)。

摘　　要：常用图像配准拼接算法中,计算复杂度会随运算数据量的增加而急剧增长,基于该特点提出了一种快速的图像配准方法.通过提取关键特征点并进行匹配实现计算数据量的降低,从而加速了图像配准.首先检测待拼接图像中的特征点及特征信息冗余区域,并根据特征点的数量调整冗余区域的大小;去除位于特征冗余区域内的特征点;使用归一化互相关及随机抽样一致性算法对剩余的关键特征点进行粗细两步匹配,求出拼接参数完成图像配准,实现图像拼接.实验结果表明,提出的方法实现了图像配准拼接,并且与改进前相比显著提高了运算速度,以lena图为例,运行时间仅为改进前的30.47%.Computational complexity grows rapidly with the increase of computation data in commonly used registration methods. A fast image registration method is proposed according to this property. The computation data is decreased and the image registration speed is improved by extracting those key feature points. Firstly, the feature points and feature redundant regions are extracted. Feature redundant regions are resized according to the number of feature points. And then the feature points in feature redundant regions are removed and the remaining feature points are regarded as key feature points. They are registered by NCC and RANSAC algorithms to calculate registration parameters and then images are stitched together according the parameters. The experimental results show that the proposed method can significantly improve the speed of image registration and mosaicking. For example, time consuming is only 30.47% of the time spent before optimization for lena diagram.

关 键 词：图像快速配准 提取关键特征点 关键特征点匹配 归一化互相关 随机抽样一致性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]