Finding Junctions Using the Image Gradient

Abstract

Junctions, the intersection points of three or more intensity surfaces in an image, are useful low-level features in machine vision, with applications in recognition, motion, grouping, and 3D line interpretation. The popular edge detectors in use today, however, such as the Laplacian of the Gaussian and the second directional derivative, fragment edges at junctions, leaving these important features undetected. This paper analyzes why edges are fragmented at junctions by differential edge operators and proposes a method for detecting junctions based on this analysis. The analysis of edge fragmentation focuses on the properties of the gradient and zero crossings of the Laplacian and the second directional derivative operators. Fragmentation is caused by the intrinsic pairing of zero crossings at junctions and by a destructive interference of edge gradient vectors which increases sensitivity to noise and quantization. We propose a junction detector that works by filling in gaps at junctions in edge maps. It uses the image gradient to guide extensions of disconnected edges at junctions. A new representation for the gradient, the bow tie map, is used to implement the endpoint growing rules, which include following gradient ridges and using saddle points in the gradient magnitude. We demonstrate the junction detector on real imagery. Finally, the paper discusses previous approaches to junction detection.... Machine vision, Edge detection, Junction detection, Zero crossings, Feature extraction, Image gradient.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 1991
Accession Number
ADA259570

Entities

People

  • David J. Beymer

Organizations

  • Massachusetts Institute of Technology

Tags

Communities of Interest

  • Air Platforms
  • Autonomy
  • C4I
  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Authentication
  • Change Detection
  • Computer Vision
  • Crossings
  • Detectors
  • Directional
  • Fragmentation
  • Geometry
  • Intensity
  • Lisp Programming Language
  • Object Recognition
  • Orientation (Direction)
  • Pattern Recognition
  • Recognition
  • Three Dimensional
  • Two Dimensional

Fields of Study

  • Computer science
  • Physics

Readers

  • Fluid Mechanics and Fluid Dynamics.
  • Sensor Fusion and Tracking Systems.
  • Superconducting Magnet Technology

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms