Fingerprint enhancement by shape adaptation of scale-space operators with automatic scale selection

Area of Science:

• Image Processing
• Biometrics
• Computer Vision

Background:

• Fingerprint image quality is crucial for automatic identification systems.
• Noise and discontinuities in fingerprint images hinder accurate feature extraction.
• Existing methods often struggle with varying image quality and local ridge structures.

Purpose of the Study:

• To develop robust fingerprint image processing mechanisms.
• To improve the accuracy of ridge orientation and width estimation.
• To enhance the performance of automatic fingerprint identification systems (AFIS).

Main Methods:

• Shape-adapted smoothing using second moment descriptors for local ridge structure processing.
• Automatic scale selection based on normalized derivatives for noise-adaptive smoothing.
• Definition of a ridgeness measure to guide smoothing in noisy areas.

Main Results:

• Successfully joined interrupted ridges without destroying singularities like branching points.
• Achieved continuity in directional fields and adaptive smoothing based on local noise.
• Produced reliable estimates of ridge orientation field and ridge width.
• Generated a smoothed grey-level version of the input fingerprint image.

Conclusions:

• The proposed shape-adapted smoothing and scale selection techniques offer reliable fingerprint image enhancement.
• These methods effectively resolve fine structures in clear areas while mitigating noise in fragmented regions.
• The techniques are valuable for developers of automatic fingerprint identification systems and other image processing applications.