基于边缘的角点分类和描述算法

Contour-based Corner Classification and Description

针对大部分角点算法只关注角点的位置信息,而忽略了角点分类的问题。文中提出了一种利用边缘方向对角点进行分类和描述的方法。根据角点附近边缘方向的分布情况,对构成角点的边缘方向进行二进制编码,从而为角点提供更多的约束。通过比较角点间的二进制编码,可找出相同类型的角点对。实验表明,该方法可实现图片中相同类型角点的匹配,具有较高的正确率。

Comers in image are important features which widely used in computer vision systems. But most of contour - based comer detectors only detect the locations of comers, but neglect the type of those points. In this paper, we proposed a comer classification and description method based on edge direction. The proposed method can encode the edge direction into binary stringaeeording to the edge directiondistribution near the comer, which preserves more information for the corners. By comparing the binary eodeof the corner, we can find the comerswhich belong to the same type. Experimental results show that the method can implement the matching of corners in same type witha high accuracy rate.