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

Principal components null space analysis for image and video classification.

Namrata Vaswani1, Rama Chellappa

  • 1Department of Electrical and Computer Engineering, Iowa State University, Ames 50011, USA. namrata@iastate.edu

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
|July 13, 2006
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

Deep Learning for the Detection of Corneal Perforation on Anterior-Segment Optical Coherence Tomography in Microbial Keratitis.

Bioengineering (Basel, Switzerland)·2026
Same author

Deep Learning for Detection of Corneal Perforation on Anterior Segment Optical Coherence Tomography in Microbial Keratitis.

medRxiv : the preprint server for health sciences·2026
Same author

Recovering Pulse Waves From Video Using Deep Unrolling and Deep Equilibrium Models.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same author

Innovation in geriatrics: what this series means for care.

Innovation in aging·2025
Same author

A perspective on AI implementation in medical imaging in LMICs: challenges, priorities, and strategies.

European radiology·2025
Same author

Encoding of Demographic and Anatomical Information in Chest X-Ray-Based Severe Left Ventricular Hypertrophy Classifiers.

Biomedicines·2025
Same journal

Change-Prior-Guided Unsupervised Change Detection of Heterogeneous Remote Sensing Images.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

AgonicDreamer: Enhancing Multi-View Consistency in Text-to-3D Generation via Rectified Score Distillation.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

BiCM-Prompt: Bidirectional Cross-Modal Prompt Tuning for Class-Incremental Learning on Multisource Remote Sensing Images.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

GoP-based Quality Enhancement on Video Compression.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

Align then Tensorize: Multi-Level Consistent Anchor Graph Learning for Scalable Multi-View Clustering.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
Same journal

Beyond Fidelity: Diverse Image Synthesis via Retrieval-Augmented Diffusion.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2026
See all related articles

We introduce principal component null space analysis (PCNSA), a new classification algorithm for object recognition with unequal class covariances. PCNSA effectively identifies classes by minimizing intraclass variance, outperforming existing methods in experiments.

Area of Science:

  • Computer Science
  • Machine Learning
  • Pattern Recognition

Background:

  • Classification algorithms often struggle with datasets where different classes have unequal and non-white noise covariance matrices.
  • Existing methods like Subspace Linear Discriminant Analysis (SLDA) may not optimally handle such complex data distributions.

Purpose of the Study:

  • To introduce a novel classification algorithm, Principal Component Null Space Analysis (PCNSA), designed for complex datasets with varying class covariances.
  • To evaluate the performance of PCNSA and its modification, progressive-PCNSA, against established classification techniques.

Main Methods:

  • PCNSA identifies class-specific subspaces (Approximate Null Spaces - ANS) with minimal intraclass variance within a Principal Component Analysis (PCA) space.
  • Classification is performed by minimizing a query's distance to a class mean within its respective ANS.

Related Experiment Videos

  • Progressive-PCNSA is proposed to also detect previously unseen classes.

    • Main Results:

    • Theoretical upper bounds on classification error probability for PCNSA were derived and compared to SLDA.
    • Experimental results demonstrated competitive or superior performance of PCNSA and progressive-PCNSA against PCA, SLDA, and various kernel-based methods (SVMs, kernel PCA, kernel DA, kernel SLDA).
    • Successful applications were shown in object recognition, face recognition, action retrieval, and abnormal activity detection.

    Conclusions:

    • PCNSA offers a robust approach for classification tasks with heterogeneous class covariance structures.
    • Progressive-PCNSA enhances utility by enabling the detection of novel classes.
    • The algorithm shows promise for real-world applications in image and video analysis.