Related Experiment Videos
A model-based method for the computation of fingerprints' orientation field
1Department of Automation, Tsinghua University, Beijing 100084, China. jzhou@tsinghua.edu.cn
Summary
This study introduces a robust model-based method for estimating fingerprint orientation fields, significantly improving accuracy for low-quality images and enhancing overall fingerprint recognition system performance.
Area of Science:
- Biometrics
- Computer Vision
- Image Processing
Background:
- The orientation field is a crucial global feature for automatic fingerprint recognition.
- Existing orientation field estimation algorithms often yield unsatisfactory results, particularly for low-quality fingerprint images.
Purpose of the Study:
- To propose a novel model-based method for accurate computation of the fingerprint orientation field.
- To enhance the robustness and performance of fingerprint recognition systems.
Main Methods:
- A combination model integrating a polynomial model (global description) and a point-charge model (local refinement at singular points) was established.
- A gradient-based algorithm was used for initial coarse field computation, followed by weighted approximation using the proposed model.
Main Results:
- The model-based method demonstrates robust performance across various fingerprint image qualities.
- The proposed algorithm significantly improves the accuracy of orientation field estimation, especially for poor quality images.
Conclusions:
- The developed model-based orientation field estimation method offers a significant advancement over existing techniques.
- Implementing this algorithm enhances the overall performance of automated fingerprint recognition systems.