Optimized Human Skeleton Detection in Complex Background Videos Using MediaPipe: A Progressive Image Enhancement Approach

Result code in IS VaVaI
<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F60461373%3A22340%2F25%3A43933484" target="_blank" >RIV/60461373:22340/25:43933484 - isvavai.cz</a>
Result on the web
<a href="http://dx.doi.org/10.1007/978-3-031-96380-3_30" target="_blank" >http://dx.doi.org/10.1007/978-3-031-96380-3_30</a>
DOI - Digital Object Identifier
<a href="http://dx.doi.org/10.1007/978-3-031-96380-3_30" target="_blank" >10.1007/978-3-031-96380-3_30</a>

Result language
angličtina
Original language name
Optimized Human Skeleton Detection in Complex Background Videos Using MediaPipe: A Progressive Image Enhancement Approach
Original language description
In the field of gait analysis using human skeleton detection to evaluate the rehabilitation process of patients with musculoskeletal disorders after brain surgeries, MediaPipe, as a tool, has gained popularity among developers for its efficiency, real-time performance, open source nature, and cross-platform compatibility. However, its accuracy diminishes when dealing with complex, real-world data, such as videos with intricate backgrounds. This paper analyzes the internal detection mechanisms of MediaPipe in human pose estimation and proposes an image-processing-based method to maximize its strengths by generating human-body-contour-based region of interest (ROI) from existing information to guide skeleton detection in subsequent frames. This approach mitigates the impact of complex dynamic backgrounds, showing initial improvements in detection accuracy. Additionally, the paper examines other environment-sensitive issues affecting MediaPipe’s performance, such as target brightness and scale, and provides insights into potential enhancements in these areas. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.
Czech name
—
Czech description
—

Project
—
Continuities
I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year
2025
Confidentiality
S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection
Research Perspectives on Software Engineering and Systems Design
ISBN
978-3-031-96379-7
ISSN
2367-3370
e-ISSN
2367-3389
Number of pages
8
Pages from-to
343-350
Publisher name
Springer Cham
Place of publication
Cham
Event location
Online
Event date
Oct 23, 2024
Type of event by nationality
WRD - Celosvětová akce
UT code for WoS article
—

Similar results(10)