A topographic method for fingerprint segmentation

Fingerprint image segmentation heavily influences the performance of automated fingerprint identification systems. We propose a segmentation technique based on the topographic properties of the image surface. The digital image is treated as a noisy sampling of the underlying continuous surface which is approximated by orthogonal Chebyshev polynomials. Ridge and valley regions are discriminated by the sign of the maximum normal curvature of this continuous surface. Different approximation accuracies are used along the orthogonal and parallel directions to the ridge orientation. This determines a lower sensitivity to small irregularities along the ridge direction then along the orthogonal direction, resulting in a high quality segmentation