Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Rotationally invariant pattern recognition by use of linear and nonlinear cascaded filters.

Ning Wu1, Robin D Alcock, Neil A Halliwell

  • 1Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough LE11 3TU, UK. n.wu2@lboro.ac.uk

Applied Optics
|July 28, 2005
PubMed
Summary

Nonlinear filters enhance pattern recognition, outperforming linear filters like MACE for distinguishing similar characters (E/F). Cascaded nonlinear filters improve noise tolerance and bacterial species detection.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Interactions Among Expressed MicroRNAs and mRNAs in the Early Stages of Fowl Adenovirus Aerotype 4-Infected Leghorn Male Hepatocellular Cells.

Frontiers in microbiology·2020
Same author

An Eu(III)-functionalized Sr-based metal-organic framework for fluorometric determination of Cr(III) and Cr(VI) ions.

Mikrochimica acta·2020
Same author

Neuropeptide S attenuates methamphetamine-induced stereotyped behavior in rats.

Biochemical and biophysical research communications·2020
Same author

Longitudinal magnetization dynamics in the quantum Ising ring: A Pfaffian method based on correspondence between momentum space and real space.

Physical review. E·2020
Same author

Multiple intraventricular metastases from lung adenocarcinoma with EGFR G719X mutation: a case report.

BMC pulmonary medicine·2020
Same author

A 75-year-old woman with primary antiphospholipid syndrome presenting with livedoid vasculopathy.

Dermatologic therapy·2020

Area of Science:

  • Pattern Recognition
  • Image Processing
  • Machine Learning

Background:

  • Linear filters, such as the Minimum Average Correlation Energy (MACE) filter, are optimal for noise-free signal detection.
  • Rotation-invariant character recognition presents challenges for traditional linear filters.
  • Assessing single-layer (linear/nonlinear) and multiple-layer (nonlinear) filters is crucial for advanced pattern recognition.

Purpose of the Study:

  • To evaluate the effectiveness of single-layer and multiple-layer filters for rotationally invariant and noise-tolerant pattern recognition.
  • To determine the suitability of linear versus nonlinear approaches for complex character differentiation.
  • To demonstrate the application of cascaded nonlinear filters in biological image analysis.

Main Methods:

Related Experiment Videos

  • Comparative analysis of linear (MACE) and nonlinear filters on a rotation-invariant character recognition task.
  • Implementation and evaluation of single-layer filters with and without nonlinear thresholding.
  • Development and testing of a two-layer cascaded nonlinear filter system.
  • Application of the developed filter system to differentiate bacterial species in phase-contrast microscopy images.

    • Main Results:

    • An optimized MACE filter failed to differentiate between characters E and F in a rotation-invariant manner.
    • A single optimized linear filter combined with a nonlinear threshold function successfully achieved the required differentiation.
    • Cascaded nonlinear filters demonstrated enhanced noise tolerance in the character recognition task.
    • A two-layer cascade effectively detected different bacterial species in phase-contrast microscopy images.

    Conclusions:

    • Nonlinear thresholding significantly enhances the discriminative power of linear filters for challenging pattern recognition tasks.
    • Cascaded nonlinear filter architectures offer superior noise tolerance and are effective for complex recognition problems.
    • This approach shows promise for automated analysis in biological imaging, such as bacterial species identification.