Abstract:
Accurate and reliable automatic personal identification is critical in wide range of application
domains such as National ID card, Electronic Commerce, ATMs etc. Biometrics which refers to
automatic identification of a person based on his physiological or behavioral characteristics is
inherently more reliable, in differentiating an authorized person from an imposter, than
traditional password and PIN number based methods. Among all the biometric techniques,
fingerprint based authentication is mostly used because of its reliability, low cost and ease of
integration.
The performance of automatic fingerprint identification systems relies heavily on the quality of
fingerprint images. The quality of fingerprint image effects the accurate extraction of minutiae.
The low quality image contains large number of false minutiae as compared to good quality
image. In this dissertation, a novel technique has been proposed for quality estimation of
fingerprint images. A set of statistical and frequency features has been calculated from the
fingerprint image. K-means clustering algorithm has been utilized to classify the fingerprint
image into four classes i.e. good, dry, normal and wet. It has been shown through experimental
results that the performance of minutiae based matcher is improved when the quality of
fingerprint image is incorporated in the matching stage.
