基于子空间近似对称Hausdorff分数图像匹配方法

Symmetrical partial Hausdorff image matching method on sub-space approximation

模板匹配方法,对于如旋转、尺度等变化造成的形变非常敏感,通常需要利用大量的参考模板进行匹配,因此匹配速度较慢.由于不同模板之间的畸变不是很大,故可以利用子空间近似的方法,利用特征分解得到模板集的特征子空间,然后利用特征子空间近似表示模板集.笔者考虑到子空间近似的特点,利用对称Hausdorff分数的表示方法,使得与模板相匹配图像特征向量空间和根据模板得到的特征向量空间基本一致,从而提高了匹配效率.

Template matching is sensible to the distortion of image rotation and scaling. The speed of template matching is degraded when a large number of templates is used. The sub-space image matching is used via the tiny distortion of different templates. The eigenvector-space is used to present the template group. The symmetrical partial Hausdorff algorithm with subspace approximation is presented here, with the advantage of the eigenvector space of template group, to improve the efficiency of image matching.