Abstract:

Fingerprint segmentation is an important step of automatic fingerprint identification system. The features of the fingerprint and the tradition fingerprint segmentation algorithms are studied. A new algorithm of fingerprint segmentation is proposed. Firstly, the print area is found quickly and the image is cropped. If the contrast is low, the histogram equalization is applied on the image. Then tophat transform is used on the image to compensate the nonuniform background intensity. After that, the algorithm of the iterative selforganizing data analysis technique is applied and the image is segmented by block. And the morphological image processing is adopted. The algorithm is tested on the different quality fingerprints of the FVC (Fingerprint Verification Competition) fingerprint databases. The experimental results indicate that the proposed algorithm is effective for different quality fingerprints and outperforms the tradition fingerprint segmentation algorithms.

Key words: segmentation;fingerprint;algorithm;ISODATA clustering;morphological processing