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

Space-time image sequence analysis: object tunnels and occlusion volumes.

Mirko Ristivojević1, Janusz Konrad

  • 1Department of Electrical and Computer Engineering, Boston University, Boston, MA 02215, USA. mirko@bu.edu

IEEE Transactions on Image Processing : a Publication of the IEEE Signal Processing Society
|February 17, 2006
PubMed
Summary

This study introduces a novel 3-D approach for spatiotemporal image sequence analysis, enhancing object segmentation and occlusion detection by analyzing multiple frames simultaneously. The method utilizes a variational framework and level-set methodology for improved accuracy in dynamic scenes.

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

A Lesion-aware Edge-based Graph Neural Network for Predicting Language Ability in Patients with Post-stroke Aphasia.

Machine learning in clinical neuroimaging : 7th international workshop, MLCN 2024, held in conjunction with MICCAI 2024, Marrakesh, Morocco, October 10, 2024, proceedings. MLCN (Workshop) (7th : 2024 : Marrakesh, Morocco)·2026
Same author

Cell classification with phase-imaging meta-sensors.

Optics letters·2024
Same author

A Lesion-aware Edge-based Graph Neural Network for Predicting Language Ability in Patients with Post-stroke Aphasia.

ArXiv·2024
Same author

MIO-TCD: A new benchmark dataset for vehicle classification and localization.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2018
Same author

Concussion, microvascular injury, and early tauopathy in young athletes after impact head injury and an impact concussion mouse model.

Brain : a journal of neurology·2018
Same author

Dictionary-based image reconstruction for superresolution in integrated circuit imaging.

Optics express·2015

Area of Science:

  • Computer Vision
  • Image Processing
  • Computational Geometry

Background:

  • Traditional image sequence analysis often processes frames pairwise.
  • Spatiotemporal segmentation and occlusion analysis are crucial for understanding dynamic scenes.
  • Existing methods may lack robustness in handling complex occlusions and object interactions.

Purpose of the Study:

  • To develop a unified framework for joint spatiotemporal analysis of image sequences.
  • To introduce novel concepts of "object tunnels" and "occlusion volumes" for enhanced segmentation.
  • To improve the accuracy of object segmentation and occlusion detection in dynamic environments.

Main Methods:

  • A variational framework utilizing motion information for segmentation.
  • Application of the level-set methodology for surface and volume estimation.

Related Experiment Videos

  • Development of multiphase level-set models to handle multiple objects and occlusions.

    • Main Results:

    • Successful spatiotemporal segmentation of moving objects using "object tunnels".
    • Accurate analysis of occlusion effects through the concept of "occlusion volumes".
    • Demonstrated effectiveness on both synthetic and natural image sequences.

    Conclusions:

    • The proposed 3-D variational level-set approach offers a robust method for joint spatiotemporal analysis.
    • The novel concepts of "object tunnels" and "occlusion volumes" significantly advance image sequence analysis.
    • This framework provides a powerful tool for applications requiring precise object tracking and occlusion handling.