Page 64 - Automated Fingerprint Identification Systems (AFIS)

HISTOR Y OF AUTOMATED FINGERPRINT IDENTIFICATION SYSTEMS 49

are run and the data analyzed, the manufacturer submits them to the FBI for
evaluation. Then, if successful, they receive a letter of certification from the
FBI, specifying compliance with either Appendix F or G. Almost all AFIS
and livescan acquisitions have specified one of the two EFTS IQS Appendices
since 1994.
It is important to note that while a vendor might have a letter of certifica-
tion for their livescan and compression products, the units still need to be
cleaned frequently and calibrated from time to time, and the compression rate
must be set properly. Think of the EPA ratings for gas mileage—your car might
not experience the same results as the tests. The same holds true for your live-
scan. In fact, the FBI has noted a constant creep in the compression rate of
electronic submittals, from around 15:1 (the specified compression rate) all
the way up to 20:1.

2.7.4 COMPRESSION STANDARD
Transmittal of the 14 fingerprint images associated with a fingerprint card (or
livescan capture) at 500ppi (in both the X and Y axis) required 10MB per
person. With 28.8kbps modems in use, there was no way they would all make
it from local police to the state identification bureau, where the FBI would
provide a wideband network (the CJIS WAN). The JPEG format, being based
on 8-X8 pixel tiles, was not compatible with fingerprint images, resulting in
banding effects upon reconstruction. As a result of this deficiency, Tom Hopper
settled on Wavelet Scalar Quantization (WSQ) compression with a compression
rate of 20:1. It was specified in the Wavelet Scalar Quantization (WSQ) Grayscale
Fingerprint Image Compression Specification; the most recent version is December
19, 1997, IAFIS-IC-0110v3. A key part of the compression scheme is that the
compressed images contain not only the compressed image data but also a copy
of the Transform Table, Quantization Table, and the Huffman Table to permit
decompression.
In the summer of 1993, the IAI challenged the FBI’s use of WSQ at 20:1
even though they had not yet seen any compressed–decompressed images. The
FBI agreed to sponsor a double blind test at NIST for the IAI. The result was
presented to the FBI in January 1994 by the chair of the IAI’s AFIS Committee,
Michael Fitzpatrick of the Illinois State Police Lab. It confirmed what the IAI
had suspected. At 20:1, approximately 81% of the fingerprints had “some blur-
ring of ridge detail with some loss of pore and ridge edge information.” While
there was no loss of Galton (second-level detail), the degradation of some third-
level detail led the FBI to settle on a 15:1 average compression rate.
In 2000, with the advent of Type 14 variable density images and JPEG 2000
compression (also based on wavelets), for the first time since 1995 (the UK