Related Experiment Video
Updated: Mar 8, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.6K
A Scan-Line Forest Growing-Based Hand Segmentation Framework With Multipriority Vertex Stereo Matching for Wearable
IEEE Transactions on Cybernetics
|January 20, 2017
Summary
This study introduces a novel framework for 3-D hand segmentation in wearable devices. The method accurately segments hands by treating objects as trees in a forest, achieving high scores in segmentation accuracy.
Area of Science:
- Computer Vision
- Human-Computer Interaction
- Machine Learning
Background:
- 3-D hand gesture interaction is crucial for wearable devices.
- Accurate hand segmentation is a prerequisite for reliable gesture recognition.
Purpose of the Study:
- To develop a robust framework for 3-D hand segmentation in wearable devices.
- To improve the accuracy and efficiency of hand segmentation for gesture interaction.
Main Methods:
- A novel framework models scene objects as directed trees in a forest.
- Employs a scan-line-based forest growing method with maximum spanning forest (MSF) construction.
- Integrates disparity and hand shape constraints for accurate segmentation, using a multipriority vertex stereo matching algorithm.
Main Results:
- The proposed method achieves a hand segmentation score of approximately 96.3%.
- The accuracy of the hand segmentation results reaches about 92.9%.
- Effectively overcomes challenges like occlusion and high similarity between fingers.
Conclusions:
- The developed framework provides effective 3-D hand segmentation for wearable devices.
- The method demonstrates high performance in accuracy and score for hand segmentation.
- Enables enhanced 3-D hand gesture interaction in wearable technology.
