Low quality fingerprint image enhancement based on Gabor filter

Fingerprint enhancement proceeds by fully utilizing the intrinsic properties of original images to improve the ridges and eliminate the noises. Although Gabor filter is a classical enhancement method, its function may suffer for the low-quality images due to the unreliable orientation and frequency map estimated by conventional approach. In this paper, a new approach has been proposed to estimate these parameters in an effective way. The local image is modeled as a non-stationary signal and the technique of Short Time Fourier Transform(STFT) is applied to it. Besides, to avoid the multi-step processing and inter-dependency, we extend the theory of probability in mathematics to get the orientation and frequency simultaneously. The performance of our approach is evaluated by the verification system of NIST and FVC2004 DB1_A database. Experimental results show that our method can improve both the image quality and the accuracy of fingerprint recognition.