一种基于取向估计的关键点检测算法

主观轮廓是一种重要的视觉心理现象,体现了人类视觉惊人的感知修复能力。针对主观轮廓处理过程中提取轮廓特征的特定需求,提出了一种基于取向估计的关键点检测算法。该算法在对边缘提取和对边缘点进行取向估计的基础上,通过对取向张量进行特征值分解获得各边缘点的各向异性信息,然后通过局部极值搜索和取向差验证提取出角点和边缘独立端点,从而对这两类关键点进行了一体化的处理。实验结果表明,该算法能比较有效地提取出主观轮廓后续处理所需要的两类关键点。

A key point detection algorithm based on orientation estimation

Subjective recognition of contour figures is an intrinsic capability of human vision and an important visual psychological phenomenon.Considering the extraction of contour features during subjective contour processing,we put forward an algorithm to detect key points based on orientation estimation.The algorithm processes two kinds of key points incorporately on the basis of edge detection and edge orientation estimation.It extracts anisotropic information of edge points through eigenvalue decomposition.Then the local maximum and minimum eigenvalue are searched out and marked as key point candidates,the orientation difference is used to verdict these candidates.Experimental results showed that this approach is efficient in detecting the corners and end points in subjective contour figures.