一种基于SURF的全景图像配准算法

摘要

传统的全景图像配准多采用基于SIFT的方法,该方法数据量大、时间效率低.提出了一种基于SURF的全景图像快速配准方法.运用SU RF提取特征点,计算特征描述符；运用低时间复杂度的K-D树最近邻搜索法实现特征点快速匹配；利用RANSAC算法剔除误匹配点；最后估计出两幅全景图像的变换矩阵.测试表明:算法具有较高的时间效率和良好的鲁棒性.%Previous panoramic images registration use approaches based on SIFT which has huge data amount andlow time-efficiency. Technique for panoramic image registration is presented, SURF ( speeded up robust features) isused to detect and descript the interest points, match the interest points by using high time-efficient K-D treenearest neighbor searching method. Mismatched points is eliminated utilizing RANSAC. Transformation matrixbetween images is estimated. The experimental result shows that it has good time efficiency and excellentrobustness.