Gradient Center Tracking:A Novel Method for Edge Detection and Contour Detection
Detecting complete contours with less clutters is a very challenging task in edge detection.This paper presents a new lightweight edge detection method,Gradient Center Tracking(GCT),to detect the main contours including the boundary and the structural lines of the objects.This method tracks the center curve of contours in the gradient image and detects edges while tracking.It makes full use of the edge correlation and contour continuity to choose edge candidates,then computes the gradient intensities of the candidates to select the real edge.In this method,the intensity of the edge is redefined as the DirectionalWeighted Intensity(DWI)which helps to present the result with more complete contours and less clutters.The GCT method outperforms Canny detector and shows better results than several learning based methods.The comparison results are shown in our experiments and a typical scheme to apply the GCT method is also provided.
Edge detection Complete contours Less clutters
Yipei Su Xiaojun Wu Xiaoyou Zhou
Harbin Institute of Technology Shenzhen Graduate School,Shenzhen,China
国际会议
广州
英文
397-407
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）