基于角点特征的高精度图像配准算法

High-accuracy Image Registration Algorithm Based on Corner Feature

为了快速准确提取图像特征和匹配点对,提出了一种基于角点检测的高精度图像配准算法。该算法首先检测参考图和实时图的角点信息,然后采用双向迭代匹配算法对参考图和实时图的角点进行匹配,再根据基于仿射变换模型的RANSAC算法对候选的匹配点对进行筛选,消除错误匹配对,并估算出最佳仿射变换参数,最后用所得变换参数对实时图进行变换和重采样来实现图像配准。实验结果表明:该算法具有良好的配准精度。

To extract exact feature points and match reliable feature point pairs, a high -accuracy image registration algorithm based on comer feature is proposed firstly, which uses the Harris operator to detect the comers in reference image and temp image, and uses a two -level restricted tactic based on bidirectional match and RANSAC algorithm based on affine transform model to gain the matched comer points. Then image registration is carried out according to the affine transform model. The experiment indicates that the algorithm proposed here has good performance in accuracy.