基于多尺度曲率乘积的鲁棒图像角点检测

为了更好地进行图像角点检测,在曲率尺度空间(CSS)框架下,提出了一种基于多尺度乘积的角点检测技术,其中曲率尺度积函数被定义为各个尺度下轮廓曲率的乘积,而角点则被定义为曲率乘积的局部极值点。这种尺度积不仅能显著地增强角点曲率极值点的峰值,同时能抑制噪声影响,而且不改变角点的位置,为了说明该技术的优点,根据角点数一致性(CCN)准则证明了该技术优于其他的角点检测算法。实验结果表明,该方法不仅具备优越的检测效果,并对噪声有较强的鲁棒性。

Robust Image Corner Detection Based on Multi-scale Curvature Product

The technique of multi-scale product corner detection is proposed in the framework of curvature scale space(CSS).A scale product function is defined as the multiplication of the curvatures of the contour at each scales.Corners are constructed as the local extreme points by limiting the scale product results.The most significant property of the scale product that it has to effectively enhanced curvature extreme peaks while can suppress noise and improve localization.The detection and localization performance analysis of the scale product are performed according to consistency of corner numbers(CCN) criteria.Experiments also demonstrate that the method has the detection results of good quality and strong robustness against noise.