Jove
Visualize
Contact Us

Related Experiment Videos

Linear multilayer ICA generating hierarchical edge detectors.

Yoshitatsu Matsuda1, Kazunori Yamaguchi

  • 1Kazunori Yamaguchi Laboratory, Department of General Systems Studies, Graduate School of Arts and Sciences, University of Tokyo, Tokyo, Japan. matsuda@graco.c.u-tokyo.ac.jp

Neural Computation
|December 1, 2006
PubMed
Summary

A new linear multilayer ICA (LMICA) algorithm efficiently extracts hierarchical edge detectors from natural scenes. This method uses stochastic multidimensional scaling and local ICA to identify and separate signal features.

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

Establishment and characterization of NKMS-1, a novel mouse NK cell line.

Scientific reports·2026
Same author

Molecular and Cell Biological Characterization of Patient-Derived Head and Neck Squamous Carcinoma Cell Lines.

Cancer science·2026
Same author

Effect of introducing a pharmacist-led support system on the administration rate of vancomycin loading dose and the 0-24-hour area under the concentration-time curve.

Antimicrobial stewardship & healthcare epidemiology : ASHE·2025
Same author

FAXC depletion contributes to tumor progression via the c-MET pathway in renal cell carcinoma.

Biochemical and biophysical research communications·2025
Same author

A case of Li-Fraumeni syndrome caused by a 3.6 kb deletion in the TP53 gene suggested by additional data from the NCC Oncopanel.

Japanese journal of clinical oncology·2025
Same author

Optimal Blood Sampling Time for Area under the Concentration-Time Curve Estimation of Vancomycin by Assessing the Accuracy of Four Bayesian Software.

Biological & pharmaceutical bulletin·2024
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

Area of Science:

  • Signal processing
  • Machine learning
  • Computer vision

Background:

  • Independent Component Analysis (ICA) is crucial for signal separation.
  • Extracting meaningful features like edge detectors from natural scenes remains challenging.
  • Existing ICA methods may struggle with efficiency and hierarchical feature extraction.

Purpose of the Study:

  • To introduce a novel ICA algorithm, linear multilayer ICA (LMICA).
  • To demonstrate LMICA's capability in efficiently extracting hierarchical edge detectors.
  • To validate the algorithm's performance on large-size natural scenes.

Main Methods:

  • LMICA employs a two-phase layer approach: mapping and local-ICA.
  • The mapping phase uses stochastic multidimensional scaling to group correlated signals.

Related Experiment Videos

  • The local-ICA phase utilizes the MaxKurt algorithm for separating neighboring signal pairs.

    • Main Results:

    • LMICA successfully extracts hierarchical edge detectors from natural scenes.
    • Numerical experiments confirm the algorithm's efficiency in processing large datasets.
    • The method effectively identifies and separates signal components relevant to edge detection.

    Conclusions:

    • LMICA offers an efficient and effective approach for hierarchical feature extraction.
    • The algorithm shows promise for applications in computer vision and signal processing.
    • LMICA advances the field of ICA by enabling efficient edge detector generation.