Segmented image-plane Fourier optics: A study of a compact design for fingerprint recognition system

A critical drawback of optical fingerprint recognition systems is the impractical size of a complete system. This drawback can be overcome, if an appropriate size of sub-lenses of a microlens array is used. This thesis presents a new method for a compact fingerprint recognition system using a microlens array. This method uses the computation of a segmented Fourier Spectrum. A segmented Fourier Spectrum of a fingerprint is suitable for fingerprint recognition when a flow of an alphabetic sequence, which codes the orientation of a spatial frequency, is used to extract the uniqueness of a fingerprint image. A flow of an alphabetic sequence is used as a characteristic expression of a fingerprint image, which is defined by the dominant spatial frequency of each segmented Fourier Spectrum. The optimal size of one segmented Fourier Spectrum block is found to be 32 x 32 pixels (1.25 mm x 1.25 mm) in 256 x 256 pixels (20 mm x 20 mm). Five gray-level fingerprint images were tested. As a result, the maximum and minimum numbers of unmatched characters in a sequence flow of the same and different fingerprints were approximately 6 and 30 respectively out of a total of 6 by 8 and 8 by 8 sectors sampling the fingerprint image.