一种新的角点角度快速估计方法

在计算机视觉领域中角点特征的提取具有相当重要的地位。角点的角度是衡量角点特征的一个极为重要的指标 ,常作为阈值决定角点的取舍。文中提出一种新的角点角度估计算法 ,算法中通过对角点边缘链码采用不同的方差的高斯函数进行卷积 ,得到角点位置的偏移 ,并给出由此偏移量计算角点角度值的公式。此算法具有较小的计算量。实验证明此算法在一定场合下能有效地完成对角点角度的估计。最后给出了估计的误差分析。

A new method for estimation of corner angle

Corner feature has considerable position in the field of computer vision. The angle is one of an important feature of corner that is often used as a threshold to decide whether the corner can be utilized. A new algorithm for estimating of corner angle is proposed. The algorithm convolutes the code chain of edge with different mean square in the neighbor the corner. Therefore the different displace of corner can be obtained and the angle of corner be estimated by using this displace. The algorithm has smaller complexity of computation and can estimate corner angle accurately in proper field. The analyses of computing errors is shown.