摘要
为了更好地实现图像拼接的实时性、提高特征点匹配的效率和图像拼接的准确度,本文提出了一种基于DBSCAN(Density-Based Spatial Clustering of Applications with Noise)与互信息的图像拼接算法.首先,为了图像拼接的实时性,用ORB(Oriented FAST and Rotated BRIEF)算法快速提取特征点,在此基础上利用DBSCAN聚类算法快速构建邻接图,通过邻接图估算图像的重叠区域;然后,用二值化互信息与欧式距离方法相结合的筛选方法,实现特征点的粗匹配,该方法可以提高特征点匹配准确度,此外根据估算的重叠区域,可以提高特征点匹配的效率;最后,用改进的RANSAC算法,计算出更精确的变换矩阵,使图像拼接的结果更准确.实验证明该方法能够实时的、高效的、精准的实现图像拼接.
In order to improve the real-time and accuracy of image mosaic and improve the efficiency of feature point matching,we propose an image stitching algorithm based on DBSCAN(Density-Based Spatial Clustering of Applications with Noise)and mutual information.Firstly,the ORB(Oriented FAST and Rotated BRIEF)algorithm is used to extract feature points quickly.Secondly,the DBSCAN clustering algorithm is used to quickly construct the adjacency graph,which is used to estimate the overlapping area of the image.Then,the rough matching of feature points is realized by the combination of binarized mutual information and Euclidean distance method,which can improve the matching accuracy of feature points.In addition,according to the estimated overlapping area,the efficiency of feature point matching can be improved.Finally,the use of the improved RANSAC algorithm can calculate a more accurate transformation matrix,making the results of image stitching more accurate.The experimental results show that our method can achieve efficient and accurate image stitching in real time.
作者
张美玉
王洋洋
吴良武
秦绪佳
ZHANG Mei-yu;WANG Yang-yang;WU Liang-wu;QIN Xu-jia(School of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310032,China;Dalian Institute of Test and Control Technology,Dalian 116013,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第4期825-829,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61672463,61672462)资助。
