Characteristic preserving binarization for fingerprint image

This paper presents a new method for binarizing fingerprint image while preserving image characteristics. Firstly, a band-stop filter method is adopted to remove the noises in high frequency. At the same time, a simply-thresholded original image is passed to Gaussian low pass filter, and outlines of fingerprint region are get by applying region growing method to the filtered image. Then, background in band-pass image is filled with outlines of the fingerprint region. After these procedures, background intensity is close to zero, thus fixed threshold one is used for binarization. Our experiment results demonstrate that band-stop method can eliminate noises effectively in the fingerprint images. And binarization in outline of fingerprint can solve the problem of disturbance of the background. Segmenting foreground from background helps select threshold and keep the characteristics of fingerprint.