privilege of the authorized person. Biometrics is the science of
identifying an individual on the basis of physiological and
behavioral features[1]. Fingerprint is such a physiological
biometric and is considered to be the oldest means of personal
identification[2][14][15]. Fingerprint identification has gained
immense popularity in the fields like social security, criminal
investigation etc. due to its two very important characteristics:
uniqueness, as no two people in the world have the same
fingerprint; not even identical twins and permanence, as
fingerprints develop at the fetal stage and remain unchanged
throughout life.
The main problem with any Automated Fingerprint Identification
System is that it has to deal with poor quality images. The process
of identification further gets complicated as, these images vary in
scale, position and orientation angle. Various automatic
fingerprint matching techniques are in use which include minutiae
based, image-based and texture-based approaches[7][9]. Minutiae
based approach is most popular of the above and is used in most
of the modern fingerprint recognition systems. In this method,
fingerprint characteristics, called minutiae (e.g. ridge bifurcation,
ridge endings etc.) are extracted and stored in order to perform
identification[8][11]. Image-based approach requires the storage
of the whole fingerprint image, as the entire image is used for
template matching. These include both optical correlation based
and transform based approaches[4][5][6][16].
This paper deals with fingerprint images in transform domain,
where Walsh transform has been used. This transform is well
suited for discrete signals[3], e.g. images, when the characteristics
of the image as whole are to be analyzed. The Walsh transform of
the fingerprint image generates its spectrum in the sequency
domain. The sectorization of this spectrum is basically performed
to reduce the number of features[10], less than if all the sequency
coefficients are considered. The fingerprints of each individual
have similar sequency characteristics, even if they are affected by
noise, shifted spatially or in orientation. The set of features, called
as feature vector, hence are similar for all the fingerprints of the
same individuals and matching is thus, obtained. Figure 1 explains
the flow of the method used in this paper. The input fingerprint is
transformed into sequency domain using intermediate Walsh
transform, explained in section 2. After sectorization feature
vectors are generated and they are matched with those of stored
fingerprints. Matching gives the output which is the best match.