In this paper, an automatic scheme for the identification of fingerprint images is presented. The scheme consists of two main processes: the extraction of distinctive points only from the template fingerprint image and the detection of their corresponding ones (if they exist) on the input fingerprint image using an implementation of the Self Organizing Maps. The correspondence quality is evaluated using a proper metric, which determines the matching between the two images. The proposed scheme was tested on fingerprint image pairs subject to known and unknown transformations using the VeriFinger_Sample_Data_Base of NeuroTechnology. The overall performance for fingerprints originated from the same and different fingers was 94.12% in terms of the Equal Error Rate.