摘要
在基于特征点的图像匹配算法中,需要一些距离函数来作为相似性度量,来检测图像是否匹配。普通Hausdorff距离具有很强的抗干扰能力和容错能力,但对噪声和部分遮挡等情况一般比较敏感。本文提出一种改进的基于标准方差的Hausdorff距离,可以减少强噪声和孤立点对匹配精度和稳定性的影响,从而可以保证匹配的稳定性和有效性。
In the image registration algorithms based on the feature points, we need some distance functions as similarity measure to detect whether the images are matched. The ordinary Hausdorff distance owns the abilities of strong anti-interference and fault-tolerance, but it is sensitive to noise and partial sheltering edge. This paper puts forward an improved Hausdorff distance based on standard variance, and it can reduce the influence of matching accuracy and stability which the high noise and outlier caused, to guarantee the stability and efficiency.
