一种基于人类感知的边缘连接方法

A Perceptual Image Processing Method for Edge Linking

边缘检测中常出现断边及伪边的现象,本文针对这一问题提出了一种边缘连接的新方法:在由改进的Canny算法得到的边缘图上,基于人类感知这一特点,对边缘点进行了分类,之后运用高、低阈值得到两幅图像,以高阈值图像作为向导,在低阈值图像中找出与高阈值图像中有确定联系的点进行边缘的连接。并对连接的结果进行了短边及伪边的去除。实验结果表明,该方法对于连接一些漏掉的边缘很有效,因此可以用于区域分割和目标识别。

The broken and fault edge segments are often present in edge detection. An improved Canny arithmetic was used to extract the image edge , then classify the edge points into three types. The classification method has a reasonable correlation with psychological properties employed by human users. Two edge images undergone hysteresis: a “high image” which uses high thresholds, and a “low image” which uses low thresholds were used. The idea is to use the high image as a guide for deciding edges and from the low image to express in the output image. The results show that the method achieves moderate improvement in “gappy” edges in the sample images. This method can be used in region segmentation and target recognition.