Effect of parameter values on fingerprint filtering

Abstract

Fingerprint is presently the most significant biometric for human verification and identification. The reason being its highest degree of uniqueness, availability, durability and consistency when compared with other biometrics such as face, nose, iris, ear, palm print and signature. The use of fingerprint in human identity management spans through stages of enrolment, enhancement, feature extraction and pattern matching. The enhancement stage involves ridge segmentation, normalization, orientation estimation, frequency estimation, filtering, binarization and thinning. Filtering is the stage at which all forms of noise and contaminations introduced into the image during enrolment are removed. The removal of noise and contaminations is necessary for accurate feature extraction and pattern matching. In some of the existing fingerprint image filtering algorithms, accurate and appropriate parameter selections are essential for obtaining optimal and satisfactory results. In this research, the existing Gabor filter was modified and the values of some standard parameters were varied. Experimental study on the adequacy of the modified algorithm and its parameter values on fingerprint filtering were investigated on the standard FVC2002 fingerprint database. Comparative analysis of the obtained results with what were obtained from some existing algorithms shows satisfactory and acceptable performances of the modified algorithm.