Image analysis – Applications – Personnel identification
Reexamination Certificate
1999-08-25
2003-07-22
Chang, Jon (Department: 2623)
Image analysis
Applications
Personnel identification
C382S284000
Reexamination Certificate
active
06597802
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the field of image processing. More specifically, the invention relates to the generation of a rolled surface image representation from a sequence of partial surface images.
BACKGROUND OF THE INVENTION
In an image processing system, the first step is to image the object of interest. Often a single view of the object is not sufficient for recognition purposes. Instead, a suitably distributed image sequence comprising of different views of the object can be used to represent all the views of the object. In order to combine this sequence into a single image, special mosaicking techniques are used.
One application in which such mosaicking would be desirable is fingerprint imaging. Fingerprints have been used for automatic authentication and identification purposes for several decades. The following journal article describes examples of the state of the prior art. This reference is incorporated herein by reference in its entirety.
N. K. Ratha, S. Chen and A. K Jain, Adaptive flow orientation based feature extraction in fingerprint images, Pattern Recognition, vol. 28, no. 11, pp. 1657-1672, Nov. 1995.
As shown in
FIG. 1
, a typical automatic fingerprint identification systems consists of the above mentioned image acquisition stage (
110
) followed by feature extraction (
120
) and feature matching (
130
). The system can be either entirely local, with the fingerprint being processed on the client, or distributed, with the image or feature list sent over a network to some remote server. In either implementation, the first step (
110
) is to acquire a fingerprint image such as shown in
FIG. 2
(item
250
). There are several techniques available for sensing the fingerprint image. These include optical, capacitance, thermal and ultrasound. A typical inkless fingerprint scanner uses a prism and total frustrated internal reflection techniques to image the finger touching the prism surface. (For example see U.S. Pat. No. 3,200,701 to W. White and U.S. Pat. No. 5,467,403 to Fishbine et al.) A newer generation of inkless scanners use electrostatic or capacitive coupling techniques to sense the ridges and valleys of the finger (cf. U.S. Pat. No. 5,852,670 to Setlak et al.).
The step (
120
) in a fingerprint analysis system is to extract tightly localized minutia features.
FIG. 2
shows such features, like the ending
203
of ridge
205
or the bifurcation
201
of a ridge. These features can be represented by a list
210
which indicates the type of each minutia and its location coordinates within the image. Other information such as the local ridge orientation or a detection confidence measure could also be added to each list element. These features are then used in final step (
130
) for matching the extracted target minutiae set to some stored reference minutiae set.
During image acquisition (
110
) in a fingerprint analysis system, the images acquired are typically of one of two types. There are “rolled” fingerprint images, which are images of a finger from nail-to-nail and “dab” or plain fingerprint images, which are images of a finger as it touches the fingerprint image acquisition surface. The fingerprint dabs are good enough for verification or identification purposes. However, the rolled fingerprints have more information for matching to images of any part of the finger, and hence are preferred for the enrollment process—registering a person's identity and storing an associated unique fingerprint template.
In the traditional method of acquiring a fingerprint using ink and paper, the rolled impressions are available effortlessly through the simple process of just rolling the finger. When using an ink-less scanner for acquiring digital images directly, obtaining a rolled digital fingerprint image is not as straightforward. These scanners are set up to quickly snap single images, and do not accumulate pressure-based markings across their surface as with the inked finger rolling on paper. Digital image scanners are preferable in most applications because they are faster and less messy. Yet rolled prints contain the greatest amount of information.
Thus a method that allowed a live scan fingerprint reader to generate the equivalent of rolled impressions would be desirable. An ink-less scanner with a fast enough image acquisition system can grab multiple partial surface images while a subject rolls a finger over the surface of the scanner. Each partial surface image just shows that portion of the finger's surface which is currently in contact with the scanner. Appropriate mosaicking software would allow the integration of such a multitude of partial surface images into a single, more detailed composite image of the finger surface.
PROBLEMS WITH THE PRIOR ART
The construction of a rolled impression from a sequence of partial fingerprint images can be viewed as an image mosaicking technique. That is, it consists of registering and combining a sequence of images with a small field of view to generate an image having a larger field of view and possibly higher resolution. Most of the prior art in this area of image mosaicking is quite compute intensive. A review of the techniques in this area is available in the paper cited below. This reference is incorporated herein by reference in its entirety.
N. K Ratha, J. H. Connell and R. M. Bolle, Image Mosacing for Rolled Fingerprint Construction International Conf. On Pattern Recognition, Brisbane, Australia, 1998, pp. 1651-1653.
The following references give further background on the reconstruction specifically of rolled fingerprints using software or hardware techniques and are also incorporated by reference in their entirety:
B. H. Fishbine, G. M. Fishbine, T. D. Klein and D. E. Germann, System for Generating Rolled Fingerprint Images U.S. Pat. No. 4,933,976, June 1990.
W. J. Chilcott, Full Roll Fingerprint Apparatus U.S. Pat. No. 4,946,276, August 1990.
E. Ranalli and B. G. Broome, Fingerprint Imaging U.S. Pat. No. 5,625,448, April, 1997
Fishbine's system uses multiple dab impressions to reconstruct a large area gray-scale fingerprint image equivalent to a rolled ink image. However, it requires the processor to compute pixel variances to determine the “active area” of each successive image (that is, the portion containing the fingerprint). Only pixels from these regions in each image are used to create the composite result. Moreover, this system only really merges pixel values in a slim transition region at the trailing edge of the active area of each new impression Locating this region requires additional computation to determine the nearest sufficiently dark area within the previously defined active area. The total amount of computation required in this approach makes it undesirable for some applications. Also, the quality of the result can be adversely affected if the determination of the impression's active area or trailing edge is inaccurate.
Chilcott's system uses a linear sensor and scanning optics to build up an image one line at a time. A series of lights is turned on and off to demarcate the line along which the optical scan is currently being taken. To develop a fill rolled print the subject rolls his finger on the glass platen at the same rate as the optical scan as indicated by the lights. Alternately, a motor can rotate the finger or the camera.
Ranalli's system is similar except it uses a flat glass scanning surface instead of a curved one. While rolled image construction is not described explicitly in either patent, the situation here is fairly straightforward since each scan line is distinct. There is no need to merge information from different scans since the linear images they produce never overlap. Instead, the individual scans can just be concatenated to form the final image.
One disadvantage of these systems is that the subject must carefully roll his finger at a certain rate or else the current scan line will miss his finger. A second disadvantage is that each portion of the finger is only imaged once. This precludes using multiple observati
Bolle Rudolf Maarten
Connell Jonathan Hudson
Ratha Nalini Kanta
Chang Jon
International Business Machines Corp.
Percello Louis J.
Peterson, Jr. Charles W.
LandOfFree
System and method for generating a rolled surface... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System and method for generating a rolled surface..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System and method for generating a rolled surface... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3043944