基于B样条模型的曲线特征点检测法

提出了一种基于B样条模型的曲线特征点检测法。该方法首先用B样条函数对原始曲线进行副近，得到原始曲线的分段解析表达式，然后利用曲线曲率分布来确定其特征点。B样条曲线分段解析的特点提出了曲率估计的精度，进而使特征点检测的准确度也得到提高，同时B样曲线拟合过程在一定程度上起到了对原始曲线的平滑作用，可抑制噪声对特征点检测的影响，并且这种平滑不具有常规高斯平滑的萎缩效应。实验结果证明了本文方法的有效性。

On Detection of Prominent Point Using B-Spline Model

A method to detect prominent points of digital curves is presented. It introduces a B spline model to approximate the original curve so that the curvatures can be accurately estimated with the analytical expression and then the prominent points can be determined from the curvature distribution. Moreover the B spline approximation makes its noise insensitive to detect the prominent points, due to its smoothness in the curve fitting which does not have shrinkage effects like Gaussian function convolution. Experimental results show the validity of this method.