摘要：拐点是数字图像中的一个重要信息载体 提出一种新的拐点检测算法 该算法并非寻找连续空间中曲率的离散近似计算方法，而是源于离散曲线的外观特征，推导出离散曲线上拐点处k个点对间欧氏距离平方和局部最小这一重要性质。基于该性质，本算法首先利用Freeman链码的性质.过滤掉物体边界上明显不可能成为拐点的象素，然后在剩余的边界点中通过寻找该局部最小值定位出拐点。给出了本算法与四种著名拐点检测算法的对比实验。

-Abstract: The inflection point is a digital image of an important information carrier　put forward a new turning point detection algorithm　the algorithm does not look for continuous space discrete approximation of the curvature calculation method, but from the appearance of characteristics of discrete curve derived discrete curve the inflection point at k-points between the squared Euclidean distance and the local minimum of this important property. Based on the nature of the algorithm is the first to use Freeman chain code in nature. Filter out objects clearly can not become a turning point on the border pixel, and then in the remaining boundary points by finding out the turning point of the local minimum position. The algorithm is given inflection point detection algorithm with the four well-known comparison of experiment.