摘要：

在谱匹配方法中，谱分解耗时较长，且图像匹配方法的效果欠佳。为此，提出一种基于加权投票的图像匹配改进方法。建立特征点候选匹配之间的亲邻矩阵，每个候选匹配以一定的权重对其他候选匹配进行投票和接收投票，通过简单的数学运算和排序操作确定最优匹配。实验结果表明，引入投票权重可改善基于加权投票图像匹配方法的匹配效果，并能保持较短的运行 时间。

关键词: 图像匹配, 谱分解, 亲邻矩阵, 目标函数, 权重, 加权投票

Abstract:

Since spectral decomposition in spectral matching is time-consuming and matching results of image matching method based on weighted voting are comparatively weak, an image matching method based on improved weighted voting is proposed. Affinity matrix of candidate correspondences between feature points is built, and every candidate correspondence gives to and receives from the other candidate correspondences voting scores with a certain weight. The optimal correspondences are computed by simple mathematical and ranking operations. Experimental results show that the introduction of voting weights gains better matching results and little running time at the same time.

Key words: image matching, spectral decomposition, affinity matrix, objective function, weight, weighted voting

中图分类号: 

• TP391