Skip to main navigation Skip to search Skip to main content

Gradient center tracking: A novel method for edge detection and contour detection

  • Yipei Su
  • , Xiaojun Wu*
  • , Xiaoyou Zhou
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

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 Directional Weighted 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.

Original languageEnglish
Title of host publicationPattern Recognition and Computer Vision - First Chinese Conference, PRCV 2018, Proceedings
EditorsJian-Huang Lai, Hongbin Zha, Jie Zhou, Cheng-Lin Liu, Tieniu Tan, Nanning Zheng, Xilin Chen
PublisherSpringer Verlag
Pages397-407
Number of pages11
ISBN (Print)9783030033972
DOIs
StatePublished - 2018
Externally publishedYes
Event1st Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2018 - Guangzhou, China
Duration: 23 Nov 201826 Nov 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11256 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference1st Chinese Conference on Pattern Recognition and Computer Vision, PRCV 2018
Country/TerritoryChina
CityGuangzhou
Period23/11/1826/11/18

Keywords

  • Complete contours
  • Edge detection
  • Less clutters

Fingerprint

Dive into the research topics of 'Gradient center tracking: A novel method for edge detection and contour detection'. Together they form a unique fingerprint.

Cite this