一种基于改进SURF的图像配准方法  

Image Registration Method Based on Improved Speed-up Robust Features

作　　者：尚明姝[1] 

机构地区：[1]哈尔滨学院工学院,黑龙江哈尔滨150080

出　　处：《微电子学与计算机》2014年第2期125-128,共4页Microelectronics & Computer

基　　金：黑龙江省高等教育学会"十二五"教育科学研究重点课题(HGJXH B1110957)

摘　　要：针对经典SURF算法存在特征点冗余、运算量大的问题,提出一种基于SURF的改进算法,首先使用SURF算法提取特征点,然后建立两级特征点筛选机制从初步检测出的特征点中筛选出信息含量高、分布均匀的特征点,减少特征点数量来减轻算法的计算量,提高算法整体性能.最后采用鲁棒性较好的PROSAC算法剔除误匹配点,计算变换矩阵.实验验证了本算法性.Considering the redundant feature points, high computational complexity in image registration method based on speed-up robust features （SURF） algorithm, an improved SURF algorithm is proposed. At first it uses SURF to extract the feature points from two images. The feature points with higher entropy are extracted. By using the distance threshold to reduce the number of feature points, the computation work can be reduced. PROSAC algorithm is used to reject false matches and compute homography matrix The experiments verify the algorithm＇s efficiency and robustness.

关 键 词：图像配准 SURF 局部熵差 PROSAC 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]