摘要
针对复杂图像的快速匹配,提出基于Shi-Tomasi角点检测的特征匹配算法。依据图像的角点特征、图像灰度和位置信息,采用最大互相关函数进行相似度计算和粗匹配,用随机样本一致性算法对匹配点对进行校正并消除错误匹配。将该算法应用于实景照片拼接,实验结果表明,对存在较大色差和形变的图像,其匹配精度为97%左右,匹配精度和速度均优于传统匹配算法。
In view of the fast image registration of complicated images,a matching algorithm based on Shi-Tomasi corner detection is described.The max correlation function is used for similarity computation and rough matching according to the corner features,image gray and location information of the images.Random sample consensus method is used for revising matching points and removing false matching.The algorithm is applied to real photo stitching.Experimental results show that the matching accuracy of this algorithm is about 97% for images with chromatic aberration or deformation and outperforms traditional point matching algorithms both in matching accuracy and running speed.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2011年第6期755-758,共4页
Journal of Nanjing University of Science and Technology
基金
国家自然科学基金(60875010)
关键词
角点检测
特征匹配
角点特征
图像灰度
位置信息
互相关函数
corner detection
feature matching
graphic point features
gray-level pixel
location information
correlation function
