Efficient Method for High-Resolution Fingerprint Image Enhancement Using Deep Residual Network

Recent research on fingerprint enhancement has progressed with the development of convolutional neural networks. Existing fingerprint image enhancement algorithms aim to denoise and improve the clarity of the ridge. High-resolution fingerprint images contain three levels of features. Pores, as one of the level 3 features, have an important contribution to the identification of high-resolution fingerprint due to its abundance and strong security. In this paper, we propose an efficient high-resolution fingerprint image enhancement method based residual learning to improve details of pores and clarity of ridges. The proposed model fully exploits the hierarchical features from the poor-quality high-resolution fingerprint image by two different paths, which are the residual-learning path and low-level path. In the residua-learning path, residual blocks with wider features are used to retain effective features from preceding and current features and extract abundant high-level information of fingerprint image. In the low-level path, shallow layers are used to get fingerprint low-level feature information. The two paths are fused to combine the low-level features and high-level features to boost the reconstruction performance of fingerprint. Extensive experimental results show that the proposed method can effectively improve the accuracy, stability and performance of the recognition, and runs much faster than existing algorithms.