Browse Prior Art Database

Compression of Fingerprint Images

IP.com Disclosure Number: IPCOM000109374D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-24
Document File: 1 page(s) / 42K

Publishing Venue

IBM

Related People

Feig, E: AUTHOR [+2]

Abstract

Presented is an efficient method for compressing fingerprint images, and a corresponding fast decompression technique. The fingerprint image is viewed as a matrix of pixel values. In a preferred embodiment, the matrix dimensions are multiples of 8, say, 8M x 8N. The image is subjected to DCT processing on each 8 x 8 block followed by quantization, which involves pointwise division of the DCT output followed by rounding to the nearest integer. The scaling is adaptive, and may be different from block to block.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 77% of the total text.

Compression of Fingerprint Images

       Presented is an efficient method for compressing
fingerprint images, and a corresponding fast decompression technique.
The fingerprint image is viewed as a matrix of pixel values.  In a
preferred embodiment, the matrix dimensions are multiples of 8, say,
8M x 8N.  The image is subjected to DCT processing on each 8 x 8
block followed by quantization, which involves pointwise division of
the DCT output followed by rounding  to the nearest integer.  The
scaling is adaptive, and may be different from block to block.

      A quantization matrix Q is chosen, depending upon application
(whether visual inspection or machine recognition; the former may be
the matrix in the JPEG document, the latter may be a matrix with all
equal entries).  A series of compressions and reconstructions are
performed on each 8 x 8 block, with quantization matrices cjQ where
the constants cj are decreasing positive numbers.  After each, the
reconstructed image is compared with the original using a weighted
least squares measure.  In a preferred embodiment, this involves the
sum
                     64  m1/4k
                     S   ---- (mk - nk)2
                     k=1   4
where the summation is over all 64 pixels, and mk and nk are the
pixel values of the original and reconstructed images, respectively.
If the sum exceeds a designed threshold parameter, then the procedure
is re...