Related Experiment Video
Updated: Feb 23, 2026
06:08
Automatic Identification of Dendritic Branches and their Orientation
Published on: September 17, 2021
2.3K
Hair segmentation using adaptive threshold from edge and branch length measures.
Ian Lee1, Xian Du1, Brian Anthony1
1Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, MA, 02139, USA.
Computers in Biology and Medicine
|September 1, 2017
Summary
This study introduces a new automatic hair segmentation method for skin imaging. The technique accurately detects and removes hair, improving skin structure analysis in clinical applications.
Area of Science:
- Dermatology
- Medical Imaging
- Computer Vision
Background:
- Non-invasive imaging aids skin disease diagnosis, but hair obstructs visualization.
- Existing hair segmentation methods struggle with thin, overlapping, or low-contrast hairs on textured skin.
Purpose of the Study:
- To develop an automatic hair segmentation method for improved skin imaging.
- To enhance the accuracy of skin structure analysis by effectively removing hair artifacts.
Main Methods:
- Hair detection via top-hat transform and modified second-order Gaussian filter.
- Hair mask generation using adaptive thresholds for edge density (ED) and mean branch length (MBL).
- Refinement of the hair mask using k-NN classification for hair and skin pixels.
Main Results:
- The algorithm achieved high sensitivity (75%) and specificity (95%) on diverse skin image datasets.
- Demonstrated superior accuracy and a better true/false positive balance compared to six state-of-the-art methods.
- Validated across various illumination levels, skin colors, and imaging platforms.
Conclusions:
- The proposed method offers a robust solution for automatic hair segmentation in clinical skin imaging.
- Improved hair removal enhances the reliability of non-invasive skin monitoring and disease diagnosis.
