基于高斯曲面模型的亚像素Harris角点定位算法

Sub-pixel Harris corner detection algorithm based on Gaussian surface model

基于灰度值的Harris角点检测算法具有广泛的应用,但随着图像处理技术的发展,越来越多的场合需要将角点定位精度提高到亚像素级。提出了一种新的方法,在Harris角点初定位的基础上,通过对兴趣值矩阵进行高斯曲面拟合逼近,用曲面局部极值估算亚像素角点位置。实验结果证明该方法具有较好的鲁棒性和较快的运算速度,并能有效地得到高精度的亚像素级角点。

Harris corner detection based on gray value of image has been widely used.With the development of image processing technology,more and more applications demand to locate the corners in sub-pixel accuracy level.Based on the result of Harris method,a novel method is proposed.The algorithm of Gaussian surface fitting model subsequently is introduced to find the local maxima from CIM（image marking corners） matrix for the purpose of estimating the corners approximately in sub-pixel level.The experimental results show that the methods mentioned in the paper are of robustness and high speed,as well as the sub-pixel corners are obtained efficiently and accurately.