Fingerprint segmentation based on hidden Markov models

An important step in fingerprint recognition is segmentation. During segmentation the fingerprint image is decomposed into foreground, background and low-quality regions. The foreground is used in the recognition process, the background is ignored. The low-quality regions may or may not be used, dependent on the recognition method. Pixel features of the gray-scale image form the basis of segmentation [3]. The feature vector of each pixel is classified, the class determining the region. Most of the known methods result in a fragmented segmentation, which is removed by means of post-processing. We solve the problem of fragmented segmentation by using a hidden Markov model (HMM) for the classification. The pixel features are modelled as the output of a hidden Markov process. The HMM makes sure that the classification is consistent with the neighbourhood. The performance of HMM-based segmentation highly depends on the choice of pixel features. This paper describes the systematic evaluation of a number of pixel features. HMM-based segmentation turns out to be less fragmented than direct classification. Quantitative measures also indicate improvement.