oru.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Fuzzy similarity-based image processing
Örebro University, Department of Technology.
2005 (English)Doctoral thesis, monograph (Other academic)
Abstract [en]

Computer vision problems require low-level operations, e.g. noise reduction and edge detection, as well as high-level operations, e.g. object recognition and image understanding. Letting a PC carry out all computations is convenient but quite inefficient. One approach for improving the performance of the vision system is to bring as much as possible of the computationally intensive low-level operations closer to the camera using dedicated hardware devices, thus letting the PC focus on high-level tasks. In this thesis we present novel fuzzy techniques for reducing noise, determining edgeness and detecting junctions as well as stereo matching measures for color images, as building blocks of complex vision systems, e.g. for robot motion control or other industrial applications.

The noise reduction is achieved by evaluating a number of fuzzy rules, each suggesting a particular filtering output. The firing strengths of the rules correspond to the degrees of similarity found among the pixels in the local processing window. The approach for determining edgeness is based on fuzzy rules that combine the estimated gradient magnitude with information about the homogeneity in different parts of the processing window. In this way the response from false edges is suppressed. In the junction detection approach we let the intersection between fuzzy sets represent the similarity between information obtained with different window sizes. The fuzzy sets represent the possible orientations of line segments in the window and non-zero intersections of the fuzzy sets indicate the presence of line segments in the window. The number of line segments characterize the nature of the junction. For the stereo matching measures, the global similarity betwen two pixels is defined in terms of fuzzy conjunctions of local similarities (color and edgeness). The proposed techniques have been designed for hardware implementation, making use of extensive parallelism and primarily simple numerical operations. The performance is shown in a number of experiments, and the strengths and limitations of the techniques are discussed.

Place, publisher, year, edition, pages
Örebro: Örebro universitetsbibliotek , 2005. , p. 136
Series
Örebro Studies in Technology, ISSN 1650-8580 ; 18
Keyword [en]
Image analysis
Keyword [sv]
Bildanalys
National Category
Computer Vision and Robotics (Autonomous Systems)
Research subject
Industrial Measurement Technology
Identifiers
URN: urn:nbn:se:oru:diva-97ISBN: 91-7668-432-6 (print)OAI: oai:DiVA.org:oru-97DiVA, id: diva2:137503
Public defence
2005-04-07, Hörsal T, Teknikhuset, Örebro universitet, Fakultetsgatan 1, 70182, Örebro, 10:00
Opponent
Supervisors
Available from: 2005-03-17 Created: 2005-03-17 Last updated: 2018-01-13Bibliographically approved

Open Access in DiVA

No full text in DiVA

Authority records BETA

Tolt, Gustav

Search in DiVA

By author/editor
Tolt, Gustav
By organisation
Department of Technology
Computer Vision and Robotics (Autonomous Systems)

Search outside of DiVA

GoogleGoogle Scholar

isbn
urn-nbn

Altmetric score

isbn
urn-nbn
Total: 1111 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf