Perceptual Organization of Linear Structures

1
Perceptual Organization of Linear Structures

• Laurent Alquier
• Computer Vision Team
• LGI2P Institute
• NIMES - France
• Email alquier_at_eerie.fr

2
Objectives

• Complete system
• Perceptual Organization of Complex Scenes
• Use of Curved Features
• Robust to noise

3
Detection of contours
4
Description

• Contour detection
• Selection of salient structures
• Extraction of Primitive Features
• Organization of hypotheses
• Application to Junctions detection and matching

For a complete description

• http//www.eerie.fr/vision/alquier/publi/po-works
  hop/

5
Contributions

• Saliency Networks
• Generic formalism
• New algorithm for optimization
• Heuristics for grouping

References
Shashua and Ullman, 1991 - Alter and Basri, 1997
6
Detection of salient groups
7
Contributions (2)

• Grouping of hypothesis and fitting
• Segments, Arcs, Points of Interest
• Graphs of relationships
• Generic framework - specialized modules
• Multi-scale

References
Mohan and Nevatia, 1992 - Sarkar and Boyer,
1993 Gao and Wong, 1993
8
Hypotheses of segments and arcs
9
Contributions (3)

• Perceptual organization of Junctions
• Structural Matching
• Temporal matching - Perceptual matching

References
Matas and Kittler, 1993 - Chang and Aggarwal, 1997
10
Detection of junctions
11
Current Results

• Stable results in cluttered images
• Generic parameters
• Reasonable time on typical computers
• Qualitative results only

12
Application to satellite imaging
13
Perspectives

• Multi-scale detection
• Quantitative evaluation of results
• Validation of hypotheses
• Hierarchical matching

Additional information
http//www.eerie.fr/vision/alquier/index.html