Fingerprint Image Enhancement Method Based on U-Net Model

Fingerprint image enhancement is a vital image processing technology that finds applications in fingerprint identification, matching, and biometric authentication. Its objective is to improve the performance and accuracy of fingerprint identification systems by enhancing the quality, clarity, and contrast of fingerprint images while reducing noise. Traditional algorithms often require adjusting parameters based on individual fingerprint images, resulting in inconsistent enhancement results. To address this challenge, this paper proposes a deep learning-based approach for fingerprint image enhancement. Experimental results demonstrate the effectiveness of U-Net architecture with different depths and the impact of incorporating attention gate. Actually, a U-Net model with a depth of six achieves superior performance, surpassing other depths, and the inclusion of attention gate further enhances the results. The proposed method not only retains more fingerprint features but also produces smoother structures and reduces noise surrounding the fingerprint image. These findings contribute to advancing fingerprint image enhancement techniques, bolstering the accuracy and performance of fingerprint identification systems in applications such as forensic analysis and biometric authentication.