Fingerprint identification algorithm

Hello,
I am working on building a fast fingerprint identification algorithm.
I have some queries regarding it.



(Q1) Query related to Fingerprint restoration/enhancement

  We are working on the development of algorithms for the Fingerprint
Authentication System.
     We use an Optical sensor to capture the fingerprints, which tends
to produce pretty good  (as against those found in most standard
databases) images. Our algorithms to discover singular points and
minutia seem to work decently, without any restoration/enhancement
algorithm. But in the later stage, matching the template with another,
we are feeling that we could improve the credibility of the
Authentication system by enhancing the fingerprint before analyzing
it.
       We have an enhancing algorithm involving Sobel filters and
gaussian filters, but it is proving computationally exorbitant. We
cannot afford this. My inquiry..


Is there any simpler (faster) fingerprint enhancing/ restoration
technique?


(Q2) Query related to Fingerprint template matching
         
        First, let me explain how we match two templates. 
        
        To start of, the core (one of the two singular points in the
fingerprint) and the minutia (bifurcations and terminations of the
ridge) are discovered. Then the minutia is interpreted with respect to
Core coordinate system (i.e. core as origin) to get a fairly immutable
representation of the image.
        There is a particular class of fingerprints (Arch type), which
has no core points. Now, how to represent the minutia? Or do we have
to employ a different matching procedure?




what algorithm do the present fingerprint identification system employ
so that identification is done so fast (in less than second)?????


Any reference or suggestion are welcomed,


waiting for reply
With regards
praveen