计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (14): 168-171.

• 图形图像处理 • 上一篇    下一篇

沿梯度方向的改进一阶差分亚像素边缘检测法

林  琦,俞水良   

  1. 同济大学 铁道与城市轨道交通研究院,上海 201804
  • 出版日期:2014-07-15 发布日期:2014-08-04

Improved first order differential subpixel edge detection method along gradient direction

LIN Qi, YU Shuiliang   

  1. Institute of Railway & Urban Mass Transit, Tongji University, Shanghai 201804, China
  • Online:2014-07-15 Published:2014-08-04

摘要: 提出了一种基于梯度方向的改进型一阶差分亚像素边缘检测方法。该方法的主要思想是,在曲线轮廓上,根据方位角对边缘像素点所在的小邻域沿梯度方向进行差分,再将修正值重投影至行列。理论上而言,该方法可弥补传统一阶差分法在曲线边缘处由于差分分布较平坦而导致精度较低的不足。理想图像的实验结果表明,该方法对半径测量、圆心定位的精度均小于0.05像素,这比采用其他亚像素法进行半径测量时可达到的精度有较大提高。

关键词: 亚像素, 一阶差分, 边缘检测, 梯度方向

Abstract: The first order differential subpixel edge detection method is improved along gradient direction.According to the azimuth of each pixel located on curve edge, the difference of its surrounding domain is gained along its gradient direction and the re-projection followed. Theoretically, the improved method compensates for the flaw that normal first order differential method fails to achieve a high accuracy due to the flat differential distribution along rows and columns. Experiments demonstrate that the accuracy of the proposed method reaches 0.05 pixel when used to detect circle center and radius, which shows its great advantage compared with the other methods.

Key words: subpixel, first order difference, edge detection, gradient direction