Filter Design and Performance Evaluation for Fingerprint Image Segmentation.

Area of Science:

• Biometrics
• Image Processing
• Computer Vision

Background:

• Fingerprint recognition is vital for security and daily applications like mobile phone unlocking.
• Accurate fingerprint image segmentation is crucial for reliable recognition performance.
• Segmentation errors, such as misclassifying foreground as background or vice-versa, negatively impact recognition by losing or introducing spurious features.

Purpose of the Study:

• To introduce a novel Factorized Directional Bandpass (FDB) segmentation method for enhanced fingerprint texture extraction.
• To establish a comprehensive benchmark dataset for evaluating fingerprint segmentation algorithms.
• To systematically compare the performance of the proposed FDB method against existing state-of-the-art algorithms.

Main Methods:

• Development of a novel Factorized Directional Bandpass (FDB) segmentation technique.
• Integration of the directional Hilbert transform of a Butterworth bandpass (DHBB) filter with soft-thresholding for texture analysis.
• Creation of a manually annotated ground truth segmentation dataset comprising 10,560 fingerprint images.

Main Results:

• The proposed FDB segmentation method demonstrated superior performance compared to four widely recognized fingerprint segmentation algorithms.
• The FDB method effectively extracts texture information, leading to improved segmentation accuracy.
• A significant reduction in both types of segmentation errors (false negatives and false positives) was observed with the FDB approach.

Conclusions:

• The Factorized Directional Bandpass (FDB) method represents a significant advancement in fingerprint image segmentation.
• The publicly available benchmark dataset and FDB implementation will facilitate further research and development in biometrics.
• The FDB method offers a robust solution for improving the accuracy and reliability of fingerprint recognition systems.