Robust Fingerprint Flow Chart Algorithm

Abstract

The orientation field is an important characteristic of human skin patterns and has a significant impact on the results of fingerprint identification. The methods of constructing the orientation field based on the gradient are diverse, but they are united by a high sensitivity to noise and defects that appear on the images during the formation of tracks. The article proposes a new method for constructing a stream field for digital images of fingerprints. The method allows to improve the solution of a number of key tasks of image processing, including the task of predicting the direction of lines in the area of skin folds, scars and other finger surface defects. The method relies on such approaches as image processing at the subpixel level, on cluster analysis of the field of image gradients and consists in the sequential application of several algorithms. These are image interpolation and estimation of gradient values on it, convolution of the gradient field with a given pattern for noise control at the subpixel level, selection of reference areas based on the construction of local quality estimates of the pattern, prediction of the direction field from reference areas over the entire image with the adaptation of the predicted values for the measurement results. The results of the proposed method were verified using the web framework created on the basis of the University of Bologna in Italy. The new verification results are compared with the verification results of the earlier method developed by the authors, and with other published algorithms on the same web framework.