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

Spatiotemporal video segmentation based on graphical models.

Yang Wang1, Kia-Fock Loe, Tele Tan

  • 1Institute for Infocomm Research, Singapore 119613. yang.wang@ieee.org

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
|July 21, 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

Bimodal EEG-fNIRS and Deep Learning for Classifying Intensity-Dependent Cortical Auditory Evoked Responses.

IEEE journal of biomedical and health informatics·2026
Same author

A scoping review of current approaches to strengths-based transition practices for autistic adolescents.

Autism : the international journal of research and practice·2025
Same author

Using Inertial Measurement Units and Machine Learning to Classify Body Positions of Adults in a Hospital Bed.

Sensors (Basel, Switzerland)·2025
Same author

Author Correction: Analysis and visualisation of electronic health records data to identify undiagnosed patients with rare genetic diseases.

Scientific reports·2024
Same author

Cluster analysis and visualisation of electronic health records data to identify undiagnosed patients with rare genetic diseases.

Scientific reports·2024
Same author

Improving Text-Independent Forced Alignment to Support Speech-Language Pathologists with Phonetic Transcription.

Sensors (Basel, Switzerland)·2023

This study introduces a novel probabilistic framework for video segmentation, unifying motion, boundary, and spatial data using graphical models. The method achieves robust, spatiotemporally coherent video segmentation results.

Area of Science:

  • Computer Vision
  • Image Processing
  • Machine Learning

Background:

  • Video segmentation is crucial for analyzing video content.
  • Existing methods often focus on either motion or region-based approaches, leading to limitations.
  • Integrating diverse information sources can improve segmentation accuracy and coherence.

Purpose of the Study:

  • To propose a unified probabilistic framework for spatiotemporal video segmentation.
  • To leverage motion, boundary, and spatial connectivity information for enhanced segmentation.
  • To develop a method that balances the strengths of motion-based and region-merging techniques.

Main Methods:

  • Utilized graphical models, specifically a Bayesian network, to integrate motion vector fields, intensity segmentation fields, and video segmentation fields.

Related Experiment Videos

  • Employed Markov random fields to ensure spatial continuity of segmented regions.
  • Maximized conditional joint probability density iteratively and used distance transformation for boundary information utilization.

    • Main Results:

    • The proposed framework generates spatiotemporally coherent video segmentation.
    • Experimental results demonstrate robustness in segmentation.
    • The approach effectively combines motion and region-based segmentation strategies.

    Conclusions:

    • The probabilistic framework offers a unified approach to video segmentation.
    • The method achieves improved segmentation quality by integrating multiple data types.
    • This work presents a balanced compromise between existing motion-based and region-merging video segmentation techniques.