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

A modified K-means algorithm for circular invariant clustering.

Dimitrios Charalampidis1

  • 1Department of Electrical Engineering, University of New Orleans, 2000 Lakeshore Dr., New Orleans, LA 70148, USA. dcharala@uno.edu

IEEE Transactions on Pattern Analysis and Machine Intelligence
|December 17, 2005
PubMed
Summary
This summary is machine-generated.

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

Saddle aortic bifurcation and iliacofemoral arteries tandem embolism in a child with dilated cardiomyopathy.

VASA. Zeitschrift fur Gefasskrankheiten·2015
Same author

Open conversion after endovascular aortic aneurysm repair with the Ovation Prime™ endograft.

The International journal of artificial organs·2014
Same author

Clinical outcomes after crossed-limb vs. conventional endograft configuration in endovascular AAA repair.

Journal of endovascular therapy : an official journal of the International Society of Endovascular Specialists·2013
Same author

Single-trocar transumbilical laparoscopy-assisted management of complicated jejunal diverticula.

Surgical laparoscopy, endoscopy & percutaneous techniques·2013
Same author

Oxidative stress contributes to methamphetamine-induced left ventricular dysfunction.

Cardiovascular research·2010
Same author

Steerable weighted median filters.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2009

This study introduces Circular K-means (CK-means), a novel algorithm for clustering directional feature vectors. CK-means offers rotation-invariant pattern recognition, outperforming traditional methods in textural image analysis.

Area of Science:

  • Computer Science
  • Pattern Recognition
  • Machine Learning

Background:

  • Pattern recognition relies heavily on feature vector extraction and clustering.
  • Directional patterns are often represented by rotation-variant vectors (Fd).
  • Achieving rotation invariance is crucial for many pattern recognition tasks.

Purpose of the Study:

  • To introduce a rotation-invariant distance measure and clustering algorithm for directional feature vectors.
  • To enhance pattern recognition by enabling invariance to pattern rotation.
  • To improve clustering accuracy and efficiency for directional data.

Main Methods:

  • Developed Circular K-means (CK-means), a K-means variant using a circular-shift invariant distance measure.
  • Proposed an efficient Fourier domain representation to reduce computational complexity.

Related Experiment Videos

  • Introduced a split and merge approach (SMCK-means) for improved clustering and cluster number estimation.

    • Main Results:

    • CK-means effectively clusters directional vectors (Fd) in a rotation-invariant manner.
    • The Fourier domain representation significantly reduces computational cost.
    • SMCK-means mitigates local minima convergence and aids in determining the optimal number of clusters.
    • Experimental results on textural images demonstrate superior performance over standard K-means with transformed features.

    Conclusions:

    • CK-means provides a robust and efficient solution for rotation-invariant clustering of directional feature vectors.
    • The proposed methods enhance the performance of pattern recognition systems dealing with directional data.
    • This work offers a significant advancement in clustering algorithms for applications requiring rotation invariance.