Adaptive binarization method for fingerprint images

An efficient method for binarization of the fingerprint images is presented in this paper. The algorithm is based on the fact that the fingerprint ridges are regions where the second directional derivative of the image is positive. The derivatives at each pixel are approximated using a facet model based on the intensity values of pixels in a certain neighborhood. It was noticed that the size of this neighborhood affects critically to the results. The size of the neighborhood is depended on the inter-ridge distance. The method based on the averaging in the direction of the ridges was used to determine inter-ridge distances. Using these inter-ridge distances, size of the neighborhood in the binarization algorithm was calculated for each pixel. Ridge directions were calculated using the gradient information of the image. The algorithm was tested using digitally acquired fingerprints. The results show that the algorithm is capable to produce very accurate binarization results.