Fingerprint Mosaicing Using Modified Phase Correlation Method

Abstract

Fingerprint is widely used biometric traits to identify individual due to uniqueness. However, the fingerprint identification is still a challenging task in forensic science for criminal investigation and identification due to less area of region of interest (i.e. ridges and valleys) in the obtained fingerprints. Entire fingerprint is obtained only when it is given intentionally. In the case of criminal incidents, it is not always possible to get entire fingerprints. In such cases, obtained fingerprints are often partial i.e. having less region of interest. Therefore, if it is possible to get more than one partial fingerprint of same finger, these can be combined to increase the area of interest using mosaicing process. This large fingerprint is used to compare with stored fingerprint database for identification. The fingerprint mosaicing is the process of joining or stitching two or more than two partial fingerprint images and create a large view of fingerprint region containing ridges and valleys. The process contains mainly three steps: 1. Image registration, 2. Matching point extraction and 3. Image stitching. The thesis proposes a fingerprint mosaicing algorithm using conventional phase correlation method with some modification. The method is a registration method which estimates only the translational and rotational parameters involved in the input images. The method also finds the single matching point in both images which is used to stitch both images. The method uses Fourier phase shift property to estimate the translational and rotational parameters involved in two partial fingerprints having overlapping region. The conventional method has some drawbacks like the method can work successfully if and only if when the overlapping region is in the leftmost top corner in one of the two fingerprints. However, it does not always happen in partial fingerprints obtained in forensic science. There are total six different possible positions of overlapping region in mosaiced fingerprint. The second drawback is that the method depends on the sequence of input, if the sequence is changed the output will also change and generate incorrect mosaiced finger-print. As fingerprint images have only grey and white curvature lines (ridges and valleys), it not possible to predict the sequence of inputs by observing images. The last drawback is that the method generates mosaiced fingerprint though the input fingerprints do not have overlapping common region, thus method is unable to check the correctness of the generated output mosaiced fingerprint. The conventional method has some drawbacks. The thesis proposes a modified phase correlation method which can solve all these limitation of the previous conventional phase correlation method and make it more robust and efficient to be used practically.