computing
displacement
Computation of image recovery
Reconstructed image
(<0.2 second on a Pentium MMX)
Reconstructing a whole image using the FC15A140
FC15A140
11/16
Reconstruction routine
The source code (written in C) of a simple routine to recon-
struct images is provided with the evaluation boards (ask TCS
for information).
This is a sample routine, and may be greatly enhanced, for
instance using image processing routines such as compres-
sion. Generally, compression algorithms search for the same
pattern in two different images, so this is exactly what we need
for reconstruction. Another way to perform that task is to make
a convolution (that is multiply the result of a Fourier transform
on each image): the distance between the two images will be
given in the phase information.
After computing the distance between the slices of the finger-
print, the complete image is created. The routine simply makes
a mean value of the superimposed pixels. From a theoretical
point of view, as you have several of the same pieces of the fin-
gerprint in different images, it is possible to enhance the con-
trast and to filter the relevant information for minutia extraction,
if needed.
Start and stop acquisition
When the user is asked to sweep his/her finger, the system will
begin to analyse incoming images to detect the finger and to
avoid storing unuseful images (in the case of storing images
for later reconstruction, if not done on the fly).
The same problem occurs when to decide to stop acquisition.
The provided source code also contains an simple analysis of
the image, using a Fast Fourier Transform. The routine detects
if there is enough energy in the spatial frequency domain
related to fingerprints and so indicates if there is a fingerprint
on the device. This analysis is done on the fly.
The best solution would be to have a processor fast enough to
perform direct analysis of the slices of fingerprint, avoiding a
real reconstruction, working only on a reduced set of useful
data.