基于中心对称特征耦合仿射度量模型的图像匹配方法

Image Matching Method Based on Variance Constraint Coupled Geometric Invariance

针对当下较多图像匹配算法利用距离度量法来实现特征匹配,没考虑图像仿射变换的影响,使得匹配结果准确率不高的问题,提出了中心对称特征耦合仿射度量模型的图像匹配算法。首先,引入Forstner算子,利用像素点的GS梯度特征提取特征点;然后,利用图像的Haar小波信息与中心对称像素点的灰度值,求取特征向量;接着,利用特征点之间的旋转、平移以及缩放的仿射特征,构造仿射度量模型,利用其计算出匹配的特征点对;最后,采用结构相似度(SSIM)函数,计算匹配点对的结构相似性,对匹配点对去伪求真,以求取最优匹配效果。实验结果表明,与当下匹配方法相比,所提算法不仅能更准确地实现图像匹配,而且还能够更好地适应具有仿射变换关系图像之间的匹配。

In response to the current problem of many image matching algorithms using distance measurement to achieve feature matching without considering the impact of image affine transformation,resulting in low accuracy of matching results,this paper proposes an image matching algorithm based on the centrosymmetric feature coupled affine measurement model.Firstly,the Forstner operator is introduced to extract feature points using the Robert gradient features of pixels;then,using the Haar wavelet information of the image and the grayscale values of the centrosymmetric pixels,the feature vector is obtained;next,using the affine features of rotation,translation,and scaling between feature points,an affine metric model is constructed to calculate the matched feature point pairs;finally,the structural similarity index measurement(SSIM)function is used to calculate the structural similarity of matching point pairs,and to remove artifacts and truth from the matching point pairs in order to obtain the optimal matching effect.The experimental results show that compared with current matching methods,the proposed algorithm can not only achieve more accurate image matching,but also better adapt to the matching between images with affine transformation relationships.