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

Sequential kernel density approximation and its application to real-time visual tracking.

Bohyung Han1, Dorin Comaniciu, Ying Zhu

  • 1Advance Project Center, Mobileye Vision Technologies, Princeton, NJ 08542, USA. bhhan@cs.umd.edu

IEEE Transactions on Pattern Analysis and Machine Intelligence
|June 14, 2008
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

Multi-plane vision transformer for hemorrhage classification using axial and sagittal MRI data.

Scientific reports·2026
Same author

Deep learning based 3D brain metastasis synthesis with configurable parameters for 3D data augmentation.

Scientific reports·2026
Same author

Single shot full plan deep learning dose computation for radiation therapy using spherical harmonics.

Medical physics·2026
Same author

Towards a cardiovascular magnetic resonance foundation model for multi-task cardiac image analysis.

Journal of cardiovascular magnetic resonance : official journal of the Society for Cardiovascular Magnetic Resonance·2025
Same author

Deep learning-based identification of patients at increased risk of cancer using routine laboratory markers.

Scientific reports·2025
Same author

Deep Learning-based Unsupervised Domain Adaptation via a Unified Model for Prostate Lesion Detection Using Multisite Biparametric MRI Datasets.

Radiology. Artificial intelligence·2024
Same journal

HardFlow: Hard-Constrained Sampling for Flow-Matching Models Via Trajectory Optimization.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Industrial Brain: Self-Evolving Neuro-Symbolic Autonomy with Causal Resilience for Cyber-Physical Systems.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Adaptive Hardness-Driven Dictionary Distillation for Incomplete Streaming View Clustering.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Mixture of Global and Local Experts with Diffusion Transformer for Controllable Face Generation.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads.

IEEE transactions on pattern analysis and machine intelligence·2026
Same journal

Achieving Text-based Person Retrieval with Any Granularity.

IEEE transactions on pattern analysis and machine intelligence·2026
See all related articles

This study introduces a novel kernel density approximation for computer vision, offering a flexible and memory-efficient way to model visual features in real-time applications.

Area of Science:

  • Computer Vision
  • Machine Learning
  • Statistical Modeling

Background:

  • Traditional probability density functions (e.g., Gaussian mixtures, kernel density estimation) in computer vision face limitations like inflexibility or high memory usage.
  • Real-time applications demand density functions that can be efficiently updated with new data, posing a challenge for existing methods.

Purpose of the Study:

  • To develop a novel kernel density approximation technique that is both memory-efficient and flexible.
  • To enable efficient sequential updates of density modes for real-time visual tracking.
  • To address the limitations of existing density modeling methods in computer vision.

Main Methods:

  • A new kernel density approximation technique is proposed, leveraging the mean-shift mode finding algorithm.

Related Experiment Videos

  • An efficient method for sequentially propagating density modes over time is described.
  • The technique allows for a variable number of density components, combining benefits of mixture and non-parametric models.

    • Main Results:

    • The proposed sequential kernel density approximation is shown to be memory-efficient while maintaining flexibility.
    • Simulations and experiments validate the accuracy and compactness of the technique.
    • The method demonstrates effective on-line target appearance modeling for visual tracking.

    Conclusions:

    • The novel sequential kernel density approximation offers a robust and efficient solution for visual feature modeling in computer vision.
    • This technique is particularly well-suited for real-time applications requiring dynamic density updates, such as visual tracking.
    • The approach successfully balances memory efficiency with the flexibility of non-parametric methods.