Fingerprint verification method employing plural correlation judgement levels and sequential judgement stages

• US 5,040,223 A
• Filed: 02/17/1989
• Issued: 08/13/1991
• Est. Priority Date: 02/17/1988
• Status: Expired due to Fees

• Current Assignee: Nippondenso Co., Ltd. (DENSO Corporation)
• Original Assignee: Nippondenso Co., Ltd. (DENSO Corporation)
• Inventors: Kawasaki, Kouzi, Kawai, Kazuyori, Nagura, Michinaga, Kamiya, Toshiharu, Eguchi, Osamu
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US07/311,834
• Time in Patent Office: 907 Days
• Field of Search: 382/2, 382/4, 382/5, 382/37, 382/38, 356/71
• US Class Current: 382/127
• CPC Class Codes: G06V 40/12 Fingerprints or palmprints
• CPC Class Codes: G07C 9/257 electronically G07C9/26 takes precedence

Abstract

A method of fingerprint verification in which an input fingerprint is compared with a previously registered fingerprint, wherein the degree of correlation between a first section of the registered fingerprint and the input fingerprint is first determined by first-stage verification processing. If the correlation is higher than a first threshold value the input fingerprint is accepted, if lower than a second threshold value it is rejected, and if between these values, second-stage verification processing is executed in which a different section of the registered fingerprint is compared with the input fingerprint, and a final judgement made accordingly.

Claims

1. A method of fingerprint verification in which fingerprint regions are processed as data representing pixel arrays, each pixel selectively taking a first state corresponding to a fingerprint ridge and a second state corresponding to a fingerprint valley, comprising steps of:

(a) providing a primary window region of a registered fingerprint image of a registered individual, said primary widow region being situated in a region of said registered fingerprint image, and at least one secondary window region of said registered fingerprint image, said secondary window region being at least partially outside said primary window region;

(b) providing an input fingerprint image which is to be verified for correspondence with said registered fingerprint image;

(c) executing first-stage verification processing by finding a region of said input fingerprint image which is configured in accordance with and has maximum correlation with said primary window region of the registered fingerprint image, and measuring a degree of said correlation, wherein the degree of correlation is related to the number of mutually non-matching pixels;

(d) making a judgement of said degree of correlation based on two predetermined thresholds that partition a range of possible correlation values into three intervals, and if said degree of correlation is judged to be within a first interval, accepting said input fingerprint image as corresponding to said registered fingerprint image and terminating processing, if said degree of correlation is judged to be within a second interval, rejecting said input fingerprint image and terminating processing, and if said degree of correlation is judged to be within a third interval such that a definite decision on acceptance or rejection of said input fingerprint image is difficult to attain, proceeding to second-stage verification processing;

(e) executing said second-stage verification processing by finding a second region of said input fingerprint image which is configured in accordance with and has maximum correlation with said secondary window region of the registered fingerprint image, and measuring a degree of said correlation; and

(f) making a judgement of said degree of correlation in step (e) based on another threshold that partitions said range of possible correlation values into fourth and fifth intervals, accepting said input fingerprint image as corresponding to said registered fingerprint image if said degree of correlation of step (e) is within the fourth interval, and rejecting said input fingerprint image if said degree of correlation is within the fifth interval.

2. A method of fingerprint verification in which fingerprint regions are processed as data representing pixel arrays, each pixel selectively taking a first state corresponding to a fingerprint ridge and a second state corresponding to a fingerprint valley, comprising steps of:

(a) providing a primary window region of a registered fingerprint image of a registered individual, said primary window region being situated in a region of said registered fingerprint image, and at least one secondary window region of said registered fingerprint image, said secondary window region being at least partially outside said primary window region;

(b) providing an input fingerprint image which is to be verified for correspondence with said registered fingerprint image;

(c) executing first-level verification processing by finding a region of said input fingerprint image which is configured in accordance with and has maximum correlation with said primary window region of the registered fingerprint image, and measuring a degree of said correlation;

(d) making a judgement of said degree of correlation based on predetermined parameters and, if said degree of correlation is judged to be within a predetermined high range, accepting said input fingerprint image as corresponding to said registered fingerprint image and terminating processing, if said degree of correlation is judged to be within a predetermined low range, rejecting said input fingerprint image and terminating processing, and if said degree of correlation is judged to be within a range such that a definite decision on acceptance or rejection of said input fingerprint image is difficult to attain, proceeding to second-level verification processing;

(e) executing said second-level verification processing by finding a second region of said input fingerprint image which is configured in accordance with and has maximum correlation with said second window region of the registered fingerprint image, and measuring a degree of said correlation; and

(f) making a decision as to acceptance or rejection of said input fingerprint image, based upon said degree of correlation obtained in said step (e);

in which said step (b) of first-level verification processing comprises comparing said primary window region of the registered fingerprint image with successively different regions of said input fingerprint image in successive position shifts of predetermined magnitude, to find a primary window region of said input fingerprint image which exhibits maximum correlation with said primary window region of said registered fingerprint image.

3. A method of fingerprint verification according to claim 2, in which said step (b) of first-stage verification processing comprises:

(a) executing said successive position shifts in steps of comparatively large magnitude, to find a first amount of displacement of said primary window region of the registered fingerprint image that provides maximum correlation with a primary window region of said input fingerprint image; and

(b) executing said successive position shifts in steps of comparatively small magnitude, beginning from an condition of said primary window region of the registered fingerprint image being displaced by said first amount of displacement, to find a second amount of displacement of said primary window region of the registered fingerprint image that provides maximum correlation with a primary window region of said input fingerprint image.

4. A method of fingerprint verification according to claim 3 which said second amount of displacement of said primary window region of the registered fingerprint image is obtained as a displacement to a position relative to said input fingerprint image such that any movement from said position in either of two mutually perpendicular pixel array direction will result in a greater number of number of mutually non-corresponding pixels than a number of number of mutually non-corresponding pixels between said primary window region of the registered fingerprint image and said input fingerprint image which exists at the current position of said primary window region.

5. A method of fingerprint verification according to claim 3, in which said steps of comparatively large magnitude are of substantially greater magnitude than an array pitch of said pixels, and in which said steps of comparatively small magnitude are substantially identical in magnitude to said pixel pitch.

6. A method of fingerprint verification according to claim 2, in which said judgement as to whether correlation is within said high range is executed by counting a number of mutually non-corresponding pixels between said primary window region of the registered fingerprint image and said input fingerprint image and judging whether said number is below a first judgement threshold value, said judgement as to whether correlation is within said low range is executed by judging whether said number of number of mutually non-corresponding pixels is above a second judgement threshold value which is higher than said first judgement threshold value, and said judgement as to whether correlation is in a range which makes a definite decision difficult is executed by judging whether said number of mutually non-corresponding pixels is between said first and second judgement threshold values.

7. A method of fingerprint verification according to claim 2, in which said step (b) of first-stage verification processing comprises steps of:

(a) expressing said registered fingerprint image and said input fingerprint image as respective patterns of density distribution of pixels which are in a predetermined one of said first and second states thereof, and

(b) computing said amount of displacement of said registered fingerprint image based on differences between said respective patterns of density distribution.

8. A method of fingerprint verification according to claim 2, in which said step (b) of first-stage verification processing comprises steps of

(a) dividing said registered fingerprint image and input fingerprint image into respective rectangular arrays of window segments, utilizing rectangular coordinate axes i and j, each said window segments comprising a rectangular array of Δ

i pixels along the direction of said i axis by Δ

j pixels along the direction of said j axis; and

(b) counting respective numbers of pixels which are in said predetermined one of the first and second states within said window segments of said registered fingerprint image and said input fingerprint image, to obtain said density distribution patterns as respective numeric arrays P_R (i, j) and P_V (i, j) of values of said numbers of pixels; and

, computing respective values i_o and j_o for a position displacement function S(i_o, j_o) which produce a minimum value for said position displacement function, said position displacement function being expressed as;

space="preserve" listing-type="equation">S(i.sub.o, j.sub.o)=f{P.sub.V (i, j)-P.sub.R (i+i.sub.o, j+j.sub.o)}

and in which said amount of displacement is obtained as respective amounts of displacement S_i and S_j along said i and j axes, expressed as;

space="preserve" listing-type="equation">S.sub.i=-i.sub.o ×

Δ

i, and

space="preserve" listing-type="equation">S.sub.j =-j.sub.o ×

Δ

j.

9. A method of fingerprint verification in which fingerprint regions are processed as data representing pixel arrays, each pixel selectively taking a first state corresponding to a fingerprint ridge and a second state corresponding to a fingerprint valley, comprising steps of:

(a) providing a primary window region of a registered fingerprint image of a registered individual, said primary window region being situated in a region of said registered fingerprint image, and at least one secondary window region of said registered fingerprint image, said secondary window region being at least partially outside said primary window region, and providing an input fingerprint image which is to be verified for correspondence with said registered fingerprint image;

(b) executing a first-stage verification processing by comparing said primary window region of the registered fingerprint image with said input fingerprint image to find an amount of displacement of said primary window region of the registered fingerprint image from an initial position thereof which results in a maximum degree of correlation with a correspondingly positioned primary window region of said registered fingerprint image, wherein the degree of correlation is related to the number of non-matching pixels;

(c) judging, based on two thresholds that partition a range of possible correlation values into three intervals, whether said degree of correlation is within a first interval, and if so, accepting said input fingerprint image as corresponding to said registered fingerprint image and terminating processing, judging whether said degree of correlation is within a second interval, and if so, rejecting said input fingerprint image and terminating processing, and judging whether said degree of correlation is within a third interval, and if so, proceeding to second-stage verification processing;

(d) executing said second-stage verification processing by applying said amount of displacement found in said step (b) to shift said secondary window region of said registered fingerprint image to an initial displaced position, then comparing said secondary window region of the registered fingerprint image with said input fingerprint image to find an amount of displacement of said secondary window region of the registered fingerprint image from said initial displaced position thereof which results in a maximum degree of correlation with said input fingerprint image; and

(e) making a judgement of said degree of correlation in step (d) based on another threshold that partitions said range of possible correlation values into a fourth and fifth intervals, accepting said input fingerprint image as corresponding to said registered fingerprint image if said degree of correlation of step (d) is within the fourth interval, and rejecting said input fingerprint image if said degree of correlation is within the fifth interval.