一种基于改进Hausdorff距离的图像匹配方法

摘要

Hausdorff distance(HD) measure is one of commonly using measures for image matching. HD measure based on kernel function is proposes in the paper. Kernel density function is used to calculate the directed HD, and the final HD is get the minimum of the two direct HDs. Experimental results show that the HD measure described in this paper are more robust in image matching under occlusion and variation.%Hausdorff距离(HD)是图像匹配常用的相似性准则.针对HD在实际图像中难以匹配的问题,提出了基于核密度函数的HD改进形式,该HD是通过对各个最小距离值进行核密度函数计算,并将函数值的和的均值作为单向HD值,无向的HD值取两个单向值的最小值.实验结果表明,提出的改进HD在遮挡、局部变化情况下比其它的HD距离具有较好匹配效果.