基于角点特征的图像配准

Corner feature-based image registration

解决匹配特征、相似性测度以及匹配策略是图像配准的技术难题.运用Harris角点,检测候选匹配点;结合特征点对齐度方法选择合适的匹配点对;最后利用线性加权算法消除误匹配点对,进行图像配准.具体方法是首先计算各幅图像多个小窗口的最大兴趣值;根据迭代公式设定合适阈值,通过比较最大兴趣值和阈值大小得到合适角点;再根据角点对齐度准则,选出两幅图像的匹配点对.实验表明由于只对提取对应角点计算对齐度,提高了效率并减少计算量.

The strategy of matching characteristics,the similarity measure and matching measure are the technical problems of image registration.In order to get the solution,the method which was based on Harris corner detector,and combined with the alignment of the feature points of image registration method was proposed.First,the value of the image multiple small window greatest interest was calculated;then according to the the iterative formula,the suitable threshold was set.Finally,By comparing the maximum interest value and threshold size,the right angle point was got;criterion which was based on the alignment of the corner,was used for selecting the points in two images.The experiments show that the efficiency is improved and amount of calculation is reduced.