一种基于置信度的边缘检测方法

该文针对传统的边缘检测方法不能同时利用像素梯度的相位信息和幅度信息的缺点,提出一种基于边缘置信度的边缘检测算法.首先,由中心象素的梯度方向确定标准模板,然后将归一化的数据矢量和标准模板的相关系数的绝对值定义为边缘置信度.在确定像素局部极值点的时候利用了非极大抑制的方法;在确定边缘点的时候用了磁滞阈值处理.在两种处理过程中,将归一化的梯度幅度和边缘置信度都作为判别依据,既增加了对弱边缘的检测能力,同时,还能很好的抑制噪声的影响.

An Approach to Edge Detection Based on Confidence

Gradient magnitude and orientation can't be used together in traditional edge detection procedure.An approach for detecting edge based on confidence is given in this paper.First,a standard template is computed by the gradient orientation of the window center,then,edge confidence is the absolute value of the correlation coefficient between the normalized data and the template.For ascertaining whether the pixel is local maximum or not,the nonmaximum suppression is used and for ascertaining edge points from candidate,hysteresis shresholding is used.In these two steps,the normalized gradient magnitude and edge confidence are used as reference together,the new algorithm shows the ability of detecting weak edges,at the same time,noise is restricted.