基于直线逼近的图形角点检测技术研究

拐点是数字图像中的一个重要信息载体,本文针对具有类直线边界的图像提出一种新的拐点检测算法。该算法并非通过在连续空间内选择支撑区域计算每个点的特征值来选择角点,而是在针对特定图形特征,首先通过Freeman链码提取图像边缘;然后通过图像支撑点和连接点的概念过滤掉一些明显不为拐点的像素;最后用直线逐段来代替图像边缘曲线,而替代直线的端点则构成了图像的拐点。

Survey of Image Corner Detection Technique Based on Line Approximating

Inflexion is are important information carriers in computer vision. A new algorithm was presented here to detect corners on contour in digital image which edge resemble lines. This algorithm was not going to detect the corner by calculating every points’ specifically values, which was defined in continuous domain. But aim to this images’ special character, fist find the points on a curve by using freeman chain-code. Second, filter the points on a curve that obviously can not be corners by the definition of supportive point and connecting point. Then, use the straight lines to replace the curve part by part, the straight lines’ end points form corners.