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

Layered motion segmentation and depth ordering by tracking edges.

Paul Smith1, Tom Drummond, Roberto Cipolla

  • 1Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, UK. pas1001@eng.cam.ac.uk

IEEE Transactions on Pattern Analysis and Machine Intelligence
|September 24, 2004
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

The ENIGMA-PD-WML Pipeline: A Containerized, User-Friendly Approach for Accurate, Standardized Segmentation of White Matter Lesions in Multi-Site MRI Data.

bioRxiv : the preprint server for biology·2026
Same author

Integrating Pathophysiology, Pharmacology, and Physical Assessment: A Competency-Based Approach in a Master's Entry Nursing Program.

Nurse educator·2026
Same author

Superior petrosal sinus stenting for symptomatic superior semicircular canal dehiscence.

Journal of neurointerventional surgery·2026
Same author

The FlEye camera: Sampling the joint distribution of natural scenes and motion.

Physical review. E·2026
Same author

Retrospective cohort analysis of nitrite and nitrate levels in postmortem biological samples after suspected suicide, 2019-24.

BMJ public health·2026
Same author

K63-linked ubiquitylation of S2P-RNAPII regulates transcription in a DNAPK inter-dependent manner in response to double-strand breaks.

Nucleic acids research·2026
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 Bayesian framework for motion segmentation using edge tracking. It accurately separates moving objects in image sequences, determining depth order and handling multiple motions effectively.

Area of Science:

  • Computer Vision
  • Image Processing
  • Machine Learning

Background:

  • Motion segmentation is crucial for understanding dynamic scenes in image sequences.
  • Existing methods often struggle with accuracy, robustness, and handling multiple moving objects.

Purpose of the Study:

  • To develop a novel Bayesian framework for robust motion segmentation.
  • To accurately identify and separate different moving objects within image sequences.
  • To determine the relative depth ordering of segmented motion layers.

Main Methods:

  • Utilizes edge detection (Canny) and the Expectation-Maximization (EM) algorithm for motion model fitting.
  • Employs edge probabilities and Markov Random Field (MRF) priors for image region segmentation.
  • Implements an efficient framework for two-motion segmentation and extends it for multiple motions.

Related Experiment Videos

  • Applies the Minimum Description Length (MDL) principle for automatic layer selection.

    • Main Results:

    • Successfully segments image sequences into distinct motion layers.
    • Accurately determines the depth ordering of foreground and background objects.
    • Demonstrates robust performance across over 30 diverse image sequences with two and three motions.
    • Achieves improved accuracy and robustness by accumulating edge probabilities over multiple frames.

    Conclusions:

    • The proposed Bayesian framework offers an effective solution for motion segmentation.
    • The method provides accurate depth ordering and handles complex multi-motion scenarios.
    • The framework is adaptable and extensible for various image sequence analysis tasks.