Abstract:
In Biometrics fingerprint matching is still an open research area achievements of high accuracy with light weight algorithms is the main problem of AFIS (Automated Fingerprint Identification System). Two techniques are proposed in this article, in first technique a novel fingerprint matching scheme based on rotational invariant Zernike moments with respect to each minutiae point is presented. Zernike features are calculated in a fixed region of interest around each minutia in a fingerprint image. Normalized Euclidean distance based similarity function is used to find the similar texture regions around each minutiae point, followed by a local greedy search with relaxation of similarities is applied to calculate the final matching score. The proposed fingerprint matching scheme is tested on all the four databases of FVC 2002 and compared with state of the art minutiae based matchers. Computational complexity of proposed matching scheme is lower than conventional minutiae based matching algorithms. Experimental results show that the proposed technique has better performance on poor quality fingerprint images in terms of matching accuracy as compared to the traditional fingerprint matching methods. On FVC 2002 databases DB3a and DB4a which contains poor quality fingerprint images, proposed fingerprint matching scheme out performs the state of the art fingerprint matching techniques.
In Second technique we propose a new algorithm for detection of Partial Fingerprint Images using “Core Point to Segmented mask distance calculation”. Recognition of partial fingerprint images is a big challenge in fingerprint recognition. It is particularly useful during fingerprint acquisition as to determine whether a user needs to realign his/her finger to ensure a complete capture of fingerprint area. Complex Filter is used to detect core point of Fingerprint images. While we generate a mask from images, this mask is the foreground area of the fingerprint which is sued for segmentation. This technique is testing on FVC-2002 DB it gives extremely good results to detect partial Fingerprint Images.
