基于自顶向下分裂聚类的图像匹配算法研究

Research of image matching algorithm based on top-down split clustering

为了提高图像匹配的效果,提出一种自顶向下分裂聚类的图像匹配算法,该算法可以获得多个目标级别的对应关系的聚类,进而找到两幅图像共存的多个目标。在互k近邻图表示模型的基础上,通过团检测方法来获得图中的团,主要是利用分裂聚类的思想,并定义了一个团密度函数,根据此函数来确定分裂终止条件。根据团检测技术获得的团恢复出团内的对应关系,从而达到图像匹配的目的。实验结果表明:该算法有较好的性能,可以应用到很多图像匹配问题中。

In order to improve the effect of image matching,this paper proposed an image matching algorithm based on topdown split clustering to obtain the clustering of corresponding relation of multiple target levels. And then it found coexistent multiple targets of two images. On the basis of mutual k neighbor graph representation model,the algorithm acquired the the cluster of the figure by cluster detection method,mainly using the idea of split clustering. And this paper defined a cluster density function,which determined split termination condition. Finally,the algorithm recovered the corresponding relation within the clusters which were obtained by the cluster detection technology,so as to achieve the goal of image matching. The experimental results show that the algorithm has good performance and it can be applied to solve a lot of image matching problems.