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

A Unified Framework for Tracking Based Text Detection and Recognition from Web Videos.

Shu Tian, Xu-Cheng Yin, Ya Su

    IEEE Transactions on Pattern Analysis and Machine Intelligence
    |April 20, 2017
    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

    Effect of motivational behavioral change-integrated rehabilitation nursing on recovery and prognosis in patients with hypertension and diabetic foot ulcers.

    Medicine·2026
    Same author

    Exosomes from bipolar disorder patients induce mania- and depression-like behaviors in mice via shared and distinct neuroimmune signatures.

    Brain, behavior, and immunity·2026
    Same author

    Twist-Induced Flat Bands and Magnetic Phase Transitions in 1T-FeCl<sub>2</sub> Bilayers: A First-Principles Study.

    The journal of physical chemistry letters·2026
    Same author

    Effects of a cardiometabolic risk-reducing dietary pattern health education intervention on cardiometabolic disease risk in community-dwelling older adults in China: a cluster-randomized clinical trial.

    The journal of nutrition, health & aging·2026
    Same author

    Cordycepin alleviates nicotine withdrawal-induced anxiety <i>via</i> AMPK/CREB/BDNF pathway activation in the hippocampus of male mice.

    Drug and alcohol dependence reports·2026
    Same author

    HACC-TNF-α-VLP Nanoparticles Facilitate HSP90-Dependent Antigen Cross-Presentation to Enhance CD8⁺ T Cell and Mucosal Immunity Against Foot-and-Mouth Disease.

    International journal of nanomedicine·2026

    This study introduces a Bayesian framework for Tracking based Text Detection And Recognition (TTDAR) in web videos. The TTDAR method enhances video text extraction by integrating detection, tracking, and recognition across multiple frames.

    Area of Science:

    • Computer Vision
    • Artificial Intelligence
    • Multimedia Analysis

    Background:

    • Video text extraction is crucial for multimedia understanding and retrieval.
    • Existing methods often process frames individually or fail to integrate text detection, tracking, and recognition effectively.
    • Web video embedded captions present unique challenges for automated analysis.

    Purpose of the Study:

    • To propose a unified Bayesian-based framework for Tracking based Text Detection And Recognition (TTDAR) from web videos.
    • To improve the mining of relationships among text detection, tracking, and recognition processes.
    • To enhance the performance of embedded caption text extraction from web videos.

    Main Methods:

    • A generic Bayesian-based framework integrating text tracking, tracking-based text detection, and tracking-based text recognition.

    Related Experiment Videos

  • Text tracking is performed using a tracking-by-detection approach.
  • Tracking trajectories are refined using detection and recognition outputs, with multi-frame integration enhancing final detection and recognition.

    • Main Results:

    • The proposed TTDAR framework effectively integrates text detection, tracking, and recognition.
    • Significant improvements in text detection and recognition performance from web videos were achieved.
    • The introduction of the USTB-VidTEXT dataset facilitates further research in this area.

    Conclusions:

    • The developed Bayesian framework offers a robust solution for embedded caption text extraction in web videos.
    • Integrating tracking, detection, and recognition within a unified model leads to superior performance.
    • The USTB-VidTEXT dataset provides a valuable resource for evaluating video text analysis techniques.