Region of Interest (ROI) Extraction for Wrist Vein Imaging

This repository provides a Python implementation of the wrist Region of Interest (ROI) extraction algorithm used in our wrist vein verification system.

Contents

• wrist_roi_extraction.py - Main ROI extraction script.
• requirements.txt - Python 3.9.2 dependency list.

Sample Results (Left Wrist Example) (Click the thumbnails to enlarge)

Input Image	1. Padding (80 px)	2. Otsu Thresholding

3. Largest Contour	4. Convex Hull	5. Major Defects (P1 & P2)

6. Key Vectors & Perpendicular Line	7. Compute P7	8. Compute P8

9. Compute P9	10. Cross-Product Direction	11. Scaled ROI

12. Final ROI (64×64)

How to Use

Install required Python 3.9.2 packages:

pip install -r .\requirements.txt

Set the captured_img variable in wrist_roi_extraction.py to your test image, then run:

python .\wrist_roi_extraction.py

Limitations

1. Fixed wrist insertion direction
   • The algorithm currently assumes the wrist enters from the opening side of the imaging device (‵ㄇ‵-shaped structure). It does not support arbitrary wrist orientations.
2. Requires a completely black background
   • The algorithm relies on background–foreground separation using Otsu’s thresholding. Complex or non-uniform backgrounds may cause segmentation failure.