Optical Fingerprint Sensors: Principles, Common Issues, and Gabor Filter Preprocessing

With the reduction of cost, optical fingerprint sensors have been adopted in low‑, mid‑, and high‑end flagship devices. Their unlocking performance has been optimized and is now comparable to traditional capacitive sensors.

Both optical and capacitive fingerprint technologies capture the ridge‑valley pattern of a fingerprint, but they differ in principle. Capacitive sensors detect the electric field variation caused by the finger’s contact with a conductive electrolyte, offering small component size, low power consumption, and weak environmental constraints.

Optical sensors acquire fingerprint images by reflecting light through a short‑focus lens placed beneath a transmissive OLED screen. This approach imposes stricter requirements on screen structure, power management, and overall device design, making development more challenging.

The typical enrollment and unlocking workflow is:

Enrollment: user presses finger → LCD backlight highlights → image capture → fingerprint data recording → template storage → success report. Unlocking: user presses finger → LCD backlight highlights → image capture → fingerprint matching → template update (if needed) → success report.

Common failure scenarios for under‑display optical fingerprint sensors include:

Incomplete press (insufficient contact area).

LCD highlight failure (no highlight, partial highlight, or uneven highlight).

Off‑target matching (large discrepancy between enrolled and presented fingerprint).

Interference from soft or hard protective films.

Wet or dirty fingers, skin peeling, clothing touch, foreign objects, and strong ambient light leakage.

New users transitioning from capacitive sensors often press too lightly, causing slow or unresponsive unlocking because optical sensors require a fully illuminated fingerprint image. Older users may encounter issues such as wet hands, peeling skin, or protective films that degrade fingerprint texture, leading to recognition failures.

Mitigation strategies include guiding users to apply proper pressure, re‑enrolling fingerprints, and leveraging template learning where successful matches are used to gradually expand the stored template.

To improve image quality, a Gabor filter‑based preprocessing algorithm is proposed. Gabor filters are linear edge‑extracting operators whose frequency and orientation characteristics resemble the human visual system, making them suitable for texture analysis.

Key Gabor parameters:

λ (lambda): wavelength of the sinusoidal factor.

θ (theta): orientation of the Gabor kernel.

ψ (psi): phase offset.

σ (sigma): standard deviation of the Gaussian envelope.

γ (gamma): spatial aspect ratio.

The implementation uses Python with OpenCV, Matplotlib, and NumPy. Fixed parameter values are applied to the fingerprint images, and the resulting enhanced images show clearer ridge patterns, especially for wet or dirty fingerprints.

Experimental results demonstrate that Gabor preprocessing can partially restore texture clarity in problematic cases, thereby improving recognition robustness.

Future development directions note that while capacitive sensors remain fast, accurate, low‑cost, and secure, optical sensors offer a more seamless full‑screen design but still lag in speed and precision. Ongoing algorithmic research and hardware optimization are expected to narrow this gap.

References: [1] https://blog.csdn.net/Ibelievesunshine/article/details/105101268 [2] http://biolab.csr.unibo.it/ResearchPages/SFinGe_Samples.asp