基于K-均值聚类图像配准算法

Research on Algorithm of Image Registration Based on K-means Clustering

针对图像配准中实时性差和精度低的问题,提出了一种基于K-均值聚类的图像配准算法。该算法根据匹配点对距离和方向特征的视差约束条件,首先利用K-均值聚类对匹配点对进行预处理,剔除错误匹配点,然后利用RANSAC进行优化,实现了图像的精确配准。实验结果表明该算法不仅提高了图像配准的精确度,而且提高了图像配准的速度。

According to the problems of poor instantaneity and low accuracy in image registration, a new image registration algorithm based on K- means clustering is proposed. Based on the parallax constraints of match point on the distance and direction feature, this algorithm first preprocess the matching point pair using K-means clustering to eliminate the false matching points, and then uses RANSAC for optimization to achieve accurate image registration. Experimental results show that the algorithm not only improves the accuracy of the image registration, but also improves the speed of image registration.