JP2000057338A - Fingerprint collating device and method and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a fingerprint collating device/method which can perform the quick and accurate collation of fingerprints regardless of quality of the fingerprint images and also to provide a storage medium. SOLUTION: When an input image is fetched via a fingerprint image input part 11, a positioning part 15 positions the input image via a matched filter based on only the phase information on the input image and a reference image. Then a correlation coefficient calculating part 16 calculates a correlation coefficient in every block of both input and reference images, and a quality correction processing part 17 corrects these correlation coefficients based on the correction data which are viewed in terms of the image quality, i.e., the local streamline direction detecting performance of a fingerprint image. Then a deciding part performs its deciding operations, based on the corrected correlation coefficients.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fingerprint collation apparatus, method and recording medium for collating a fingerprint by comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance. Fingerprint matching device that performs quick and accurate fingerprint matching without being affected by the image quality of the image,
The present invention relates to a method and a storage medium.

[0002]

2. Description of the Related Art Heretofore, while absorbing local deformations (non-rigid deformations) other than translation and rotation included in a fingerprint image, individuals using the fingerprint of a person having the characteristics of being universal and lifelong invariant have been used. A fingerprint collation technology for performing authentication is known.

For example, Japanese Patent Application Laid-Open No. 9-282458 discloses a correlation between a reference effect block on a reference image corresponding to a target block of an input image and an input influence block on the input image corresponding to the target block of the input image. The block position of the target block is specified based on the value, and the matching degree between the input image and the reference image is calculated based on the correlation value between each block of the specified input image and the corresponding block of the reference image. An image matching device is disclosed.

Specifically, in this conventional technique, an input image and a reference image are divided into a plurality of blocks, and the blocks are arranged so as to absorb non-rigid deformation in the input image.
The correlation coefficient between the input image from which the non-rigid deformation has been removed and the reference image is obtained for each block, and the average value is used as the matching degree. If the degree of collation is equal to or greater than a predetermined threshold value, the user is recognized as an individual. If the collation degree is less than the predetermined threshold value, the user is identified as not an individual.

[0005]

However, if the average value of the correlation coefficient between the input image and the reference image is simply used as the degree of collation as in this prior art, the collation result is large in the image quality of the input image. The probability that the fingerprint of the input image is erroneously recognized as not the person in spite of the fact that the fingerprint is of the person (person exclusion error rate) increases.

Originally, the image quality of such an input image is
Since the quality of the input image can be easily changed depending on the dry state of the finger of the person to be authenticated, it is not appropriate to judge that the input image is not the person in spite of the fact that the image quality is not so bad.

On the other hand, in order to lower the personal exclusion error rate due to the deterioration of the image quality of the input image, simply lowering the threshold value for comparison of the matching degree is a fingerprint of another person. Nevertheless, the probability of falsely recognizing that the user is a person (an error rate of accepting another person) increases, and the existence value of the fingerprint collation device decreases.

[0008] From these facts, it is an important issue how to realize a fingerprint matching device capable of performing fingerprint matching quickly and accurately without being affected by the image quality of an input image.

Therefore, the present invention solves the above-mentioned problems,
An object of the present invention is to provide a fingerprint matching device, a fingerprint matching device, a method, and a storage medium that can quickly and accurately perform fingerprint matching without being affected by the image quality of a fingerprint image.

[0010]

In order to achieve the above object, according to the first aspect of the present invention, an input image obtained by optically reading a fingerprint is compared with a reference image registered in advance to check the fingerprint. Calculating means for calculating a correlation value between the input image and the reference image, correcting means for correcting the correlation value calculated by the calculating means based on the image quality of the input image, and the correcting means A determination unit that determines whether a user having a fingerprint included in the input image is the user who has registered the reference image based on the corrected correlation value.

According to a second aspect of the present invention, there is provided a fingerprint matching device for comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance to perform fingerprint matching. Calculating means for calculating the correlation value of the reference image; correcting means for correcting the correlation value calculated by the calculating means based on the image quality of the input image and the reference image; and correcting means for correcting the correlation value calculated by the correcting means. A determination unit that determines whether a user having a fingerprint included in the input image is the user who has registered the reference image.

According to a third aspect of the present invention, the correction means creates correction data corresponding to the image quality of the input image and the reference image from the clarity of the streamline direction of the input image and the reference image. Then, the correlation value is corrected using the correction data.

According to a fourth aspect of the present invention, there is provided a fingerprint matching method for comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance to perform fingerprint matching. The correlation value of the reference image is calculated, the calculated correlation value is corrected using the clarity of the streamline direction indicating the image quality of the input image and the reference image, and the input image is corrected based on the corrected correlation value. It is characterized in that it is determined whether or not the user having the included fingerprint is the user who registered the reference image.

According to a fifth aspect of the present invention, there is provided a recording medium used in a fingerprint collating apparatus for collating a fingerprint by comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance. Calculating the correlation value between the input image and the reference image, correcting the calculated correlation value using the clarity of the streamline direction indicating the image quality of the input image and the reference image, and based on the corrected correlation value. A program for determining whether or not the user having the fingerprint included in the input image is the user who registered the reference image.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. In the present embodiment,
When a correlation coefficient between an input fingerprint image (hereinafter, simply referred to as “input image”) obtained by reading a fingerprint and a reference image (hereinafter, simply referred to as “reference image”) obtained by reading a fingerprint in advance is used as a collation value. Is shown.

First, the relationship between the collation value of the input image and the reference image focused on by the present invention and the image quality of the input image will be described.

FIG. 2 is a diagram for explaining the relationship between the collation value of the input image and the reference image and the image quality of the input image.

When the image quality of the input image and the reference image (hereinafter collectively referred to as "fingerprint image") is good, the distribution of the collation value between the input image and the reference image is as shown in FIG. Become. Here, distribution C shows the case where the input fingerprint is the person himself,
Distribution E shows the case where the input fingerprint is another person.

As described above, when the distribution C1 when the input fingerprint is a person and the distribution E1 when the input fingerprint is another person are separated, the threshold value A is set at the position shown in FIG. Fingerprint collation can be performed.

On the other hand, when the image quality of the fingerprint image deteriorates, the distribution of the collation value between the input image and the reference image substantially matches the distribution shown in FIG. The distribution is shifted to lower values. The reason is that when the image quality of the fingerprint image is deteriorated, the correlation coefficient between the input image and the reference image, that is, the degree of collation is also reduced.

Therefore, in such a case, the threshold value B is set at the illustrated position between the distribution C2 when the input fingerprint is the person himself and the distribution E2 when the input fingerprint is another person, so that the fingerprint can be accurately obtained. Collation can be performed.

As described above, when a fingerprint image can be captured with a relatively constant image quality, the threshold value A or B is fixedly set as shown in FIGS. 2 (a) and 2 (b). By doing so, fingerprint collation can be performed with high accuracy. However, actually, since the image quality changes each time due to the degree of dryness of the finger, the distribution of FIG. The distribution is as shown in FIG.

In this case, if the threshold value A is always used, the false exclusion error rate increases, and if the threshold value B is used, the false acceptance error rate increases.

As described above, since the image quality of the fingerprint image directly affects the collation value between the input image and the reference image, control of the threshold value or correction of the collation value is performed in response to the image quality of the fingerprint image. Otherwise, the false exclusion error rate or false acceptance error rate will increase, and as a result, the fingerprint matching accuracy will decrease.

Therefore, in the fingerprint collation apparatus shown in the present embodiment, the collation value between the input image and the reference image is corrected in accordance with the image quality of the fingerprint image, thereby reducing the exclusion error rate and the false acceptance error rate. I have to do that. Instead of correcting the collation value, a threshold value can be controlled.

Next, the configuration of the fingerprint matching device used in the present embodiment will be described.

FIG. 1 is a diagram showing a configuration of a fingerprint matching device used in the present embodiment.

As shown in FIG.
Are a fingerprint image input unit 11 and a geometric distortion correction unit 12
And a reference image management unit 13, a reference image storage unit 14, a positioning unit 15, a correlation coefficient calculation unit 16, a quality correction processing unit 17, and a determination unit 18.

The correlation coefficient calculating section 16, the quality correction processing section 17, and the judging section 18 are described in claims 1 and 2, respectively.
Corresponds to calculation means, correction means and determination means.

The fingerprint image input unit 11 is an image input unit for optically reading a fingerprint of a finger by a CCD through a prism and a lens. And a part of the light diffusely reflected by the finger is detected by the CCD through the lens.
The input image or the reference image detected by the fingerprint image input unit 11 is output to the geometric distortion correction unit 12.

The geometric distortion correction unit 12 is a processing unit that corrects geometric distortion included in a fingerprint image, and specifically, is caused by the fact that the contact surface between the finger and the prism is not parallel to the CCD surface. This is a processing unit for removing distortion and the like caused by the center projection.

The geometric distortion correction unit 12
When the reference image is received from the fingerprint image input unit 11, the corrected reference image is output to the reference image management unit 13. When the input image is received, the corrected input image is sent to the positioning unit 15. Output.

The reference image management unit 13 is a management unit that manages the reference image to be compared with the input image using the reference image storage unit 14, and specifically, the geometric distortion correction unit 12
, The reference image is registered in the reference image storage unit 14, and the reference image requested by the positioning unit 15, the correlation coefficient calculation unit 16 and the quality correction processing unit 17 is stored in the reference image storage unit. 14 and output.

The positioning unit 15 is a matched filter (Symmetric) using only the phase information of the input image and the reference image.
This is a processing unit that performs position adjustment of both images using a Phase Only Matched Filter).

More specifically, the phases of the Fourier transform of the input image and the reference image are detected, the product of the complex numbers is obtained, the inverse Fourier transform is performed, and the position with the maximum value is used to perform positioning with high accuracy. Do.

The correlation coefficient calculating section 16 is a processing section that calculates a correlation coefficient C for each block of the input image and the reference image, and outputs the calculated correlation coefficient C to the quality correction processing section 17.

Specifically, the input image is f (x, y), the reference image is g (x, y), and the center of each block [i, j] of the reference image is the image origin. Gi, j (x, y) is the image after the coordinate transformation
The image obtained by performing the coordinate transformation so that the estimated corresponding position in the input image of each block becomes the image origin is defined as fi, j (x,
y), the correlation coefficient in each block is Indicated by Here, fi, j (xu, yv) is f
An image obtained by translating i, j (x, y) by (u, v), Becomes Note that W indicates a block area.

The quality correction processing unit 17 detects the quality of the fingerprint image based on the local deviation of the covariance of the input image and the reference image in the streamline direction, obtains a correction coefficient, and uses this correction coefficient to obtain a correlation coefficient. The processing unit corrects the correlation coefficient C between the input image and the reference image calculated by the calculation unit 16.

That is, the higher the quality, the more clearly the streamline direction of the fingerprint in the fingerprint image can be detected. Therefore, the quality of the fingerprint image is detected by the local deviation of the covariance of the streamline direction, and the correction coefficient and I do.

Specifically, an image is represented by f (x, y),
When the value image is f ′ (x, y), it is a spatial derivative at each pixel. Can determine the streamline direction. However, the x partial differential and the y partial differential are set such that the maximum possible value thereof is 1. The reason why the binary image is used is that the quality value is not directly affected by the gray value of the image. Here, the streamline covariance matrix S
(X, y) and its singular value decomposition are calculated by The difference between the maximum singular value and the minimum singular value increases as the bias of the covariance increases, and a strong directionality can be detected. Note that N indicates the vicinity of the pixel (x, y), h1 indicates the maximum singular value, and v1 indicates the eigenvector associated therewith. Also, h2 indicates the minimum singular value, and v2 is the eigenvector associated therewith.

Therefore, the average quality q of the fingerprint image
Is It can be considered that the larger the q, the better the quality. Here, E [x] indicates an average regarding x. For example, if the fingerprint image is random noise, the direction is not determined, and thus q is almost 0, and conversely, q is close to 1 if the directionality is strong.

The correction coefficient c of the fingerprint image is It will be obtained as. Here, the constants α and β are experimentally obtained, and are limited to c = 0 when c ≦ 0.

If the correction coefficients cf and cg (> 0) of the input image f and the reference image g are obtained, the corrected correlation coefficient C 'is obtained from the calculation formula of C' = C * cf * cg. To get.

As described above, the quality correction processing unit 17 obtains the image quality of the input image and the reference image from the local deviation of the covariance in the streamline direction, and calculates the image quality calculated by the correlation coefficient calculation unit 16. The correlation coefficient C when not considered is to be corrected.

The judging section 18 is a processing section for judging whether or not the fingerprint imprinter is the person based on the correlation coefficient C ′ corrected by the quality correction processing section 17. The correlation coefficient C ′ received from the processing unit 17 is compared with a predetermined threshold value, and if the correlation coefficient C ′ is equal to or greater than this threshold value, it is determined that the user is the person.

The configuration of the fingerprint matching device 10 used in the present embodiment has been described above.

Next, a processing procedure of the fingerprint collating apparatus 10 shown in FIG. 1 will be described. Here, it is assumed that the reference image has already been registered in the reference image storage unit 14.

FIG. 3 is a flowchart showing a processing procedure of the fingerprint collating apparatus 10 shown in FIG.

As shown in FIG.
When the input image is captured by the image input unit 11, the phase of the Fourier transform of the input image is detected (step 301), and the product of this phase and the phase of the Fourier transform detected from the reference image is calculated. (Steps 302 to 3
03), Fourier inverse transform is performed (step 304).

Then, the positions x and y where the values are maximum
Is obtained (step 305), the input image is shifted to be aligned (step 306), and the correlation coefficient of each block between the input image and the reference image is obtained (step 30).
7).

Then, the quality correction processing section 17 calculates the quality correction amounts from the local covariance bias in the streamline direction of the input image and the reference image (step 308). By multiplying the numbers (step 309), the correlation coefficient is corrected in terms of image quality.

Then, the judging section 18 compares the corrected correlation coefficient with a predetermined threshold value, and if the correlation coefficient is larger than the threshold value, judges that the user is the subject and if the correlation coefficient is small. If not, it is determined that the user is another person (step 310).

As described above, in the fingerprint collating apparatus 10,
The correlation coefficient indicating the degree of collation between the input image and the reference image is corrected from the viewpoint of image quality, and it is determined whether or not the user is a person using the corrected correlation coefficient.

Next, the relationship between streamline direction detectability and fingerprint image quality used by the quality correction processing unit 17 shown in FIG. 1 will be specifically described.

FIG. 4 is an explanatory diagram showing the relationship between streamline direction detectability and fingerprint image quality used by the quality correction processing section 17 shown in FIG.

As shown in FIG. 5A, when a fingerprint of a person is viewed locally, it can be regarded as a group of streamlines. Therefore, as shown in FIG. When an image can be captured, the direction of the streamline is clear as shown in FIG. 9C, and the difference between the maximum value and the minimum value of the likelihood of the streamline in each direction is large.

However, when the image quality of the fingerprint image is deteriorated and the fingerprint is blurred, the fingerprint cannot be regarded as a streamline as shown in FIG. As a result, the difference between the maximum value and the minimum value of the streamline likelihood in each direction disappears as shown in FIG.

From this, it can be seen that the local deviation of the covariance in the streamline direction of the fingerprint image indicates the degree of streamline cutting of the fingerprint image, that is, the image quality of the fingerprint image. For this reason, in the present embodiment, this “local deviation of the covariance in the streamline direction of the fingerprint image” is used as a correction amount that responds to image quality.

More specifically, as shown in FIG. 5, the fingerprint image is binarized to obtain the x-direction derivative and the y-direction derivative (step 501), and the filter output of the neighborhood (i, j) including the target pixel is used. A 2 × 2 covariance matrix shown is obtained (step 502), and this covariance matrix is converted into a singular value (a square root of an eigenvalue).
And the absolute value of the difference between the two singular values is obtained (step 503) and used as the correction amount.

As described above, in the present embodiment,
After performing position matching by a matched filter using only the phase information of the input image and the reference image, a correlation coefficient for each block of the input image and the reference image is obtained, and image quality such as local streamline direction detectability of the fingerprint image is obtained. Since the correlation coefficient is corrected by the correction data viewed from the viewpoint of and the determination is made based on the correlation coefficient, the following effects can be obtained.

1) Fingerprint collation can be performed quickly and accurately without being affected by the image quality of a fingerprint image.

2) Even if the image quality of the fingerprint image fluctuates, it is possible to suppress a rise in the false exclusion error rate and the false acceptance error rate, and to stably perform fingerprint collation.

In the present embodiment, a case has been described in which the local covariance in the streamline direction of the fingerprint image is used as an index of the streamline direction detectability. However, the present invention is not limited to this. Instead, other indicators of streamline direction detectability can be used.

For example, when the difference between the maximum value and the minimum value of the amplitude outputs of a number of directional filters is used as an index of streamline direction detectability, as shown in FIG. 6, a fingerprint image (binary image) , A real part R_k (x, y) and an imaginary part I_k (x, y) are obtained (step 601), and the amplitude A_k (x, y) of the filter output is calculated as A_k (x, y) = sqrt ｛R_k (x, y) × R_k (x, y) + I_k (x,
y) × I_k (x, y)} (step 60
2) The difference between the maximum value and the minimum value among the amplitudes of the n filter outputs is obtained (step 603), and can be used as correction data.

In this embodiment, for convenience of explanation, only the reference image data is stored in the reference image storage unit 14. However, the phase of the Fourier transform of the reference image calculated by the positioning unit 15 and the reference image data are stored. Since the quality correction amount of the reference image calculated by the quality correction processing unit 17 is not affected by the input image, it can be calculated in advance when registering the reference image. In this case, the phase and the quality correction amount of the reference image calculated at the time of registration of the reference image are stored in the reference image storage unit 14 together with the reference image.

In some cases, only a high-quality image may be registered when the reference image is registered. In this case, the correction is performed based only on the image quality of the input image.

[0067]

As described in detail above, the present invention provides
A correlation value between the input image and the reference image is calculated, the calculated correlation value is corrected using the clarity of the streamline direction indicating the image quality of the input image and the reference image, and the input image is corrected based on the corrected correlation value. Is configured to determine whether or not the user having the fingerprint included in the reference image is the user who registered the reference image, the following effects can be obtained.

1) Fingerprint collation can be performed quickly and accurately without being affected by the image quality of a fingerprint image.

2) Even if the image quality of the fingerprint image fluctuates, it is possible to suppress a rise in the false exclusion error rate and the false acceptance error rate, and to stably perform fingerprint collation.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration of a fingerprint matching device used in the present embodiment.

FIG. 2 is a diagram for explaining a relationship between a collation value of an input image and a reference image and image quality of the input image.

FIG. 3 is a flowchart showing a processing procedure of the fingerprint matching device shown in FIG. 1;

FIG. 4 is an explanatory diagram showing a relationship between streamline direction detectability and fingerprint image quality used by the quality correction processing unit shown in FIG. 1;

FIG. 5 is a flowchart illustrating a procedure for creating correction data in a case where local deviation of covariance in a streamline direction of a fingerprint image is used as streamline direction detectability.

FIG. 6 is a flowchart showing a procedure for creating correction data when the difference between the maximum value and the minimum value of the amplitude outputs of a number of directional filters is used as the streamline direction detectability.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 10 ... Fingerprint collation apparatus 11 ... Fingerprint image input part 12 ... Geometric distortion correction part 13 ... Reference image management part 14 ... Reference image storage part 15 ... Alignment part 16 ... Correlation coefficient calculation part 17 ... Quality correction processing part 18 ... Judgment unit

Claims (5)

[Claims] 1. A fingerprint matching device for comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance and performing fingerprint matching, wherein a calculating means for calculating a correlation value between the input image and the reference image. Correction means for correcting the correlation value calculated by the calculation means based on the image quality of the input image; and a user having a fingerprint included in the input image based on the correlation value corrected by the correction means. A fingerprint matching device comprising: a determination unit configured to determine whether or not the user is a user who has registered the reference image. 2. A fingerprint matching device that compares an input image obtained by optically reading a fingerprint with a reference image registered in advance and performs fingerprint matching, wherein a calculating unit calculates a correlation value between the input image and the reference image. A correction unit that corrects the correlation value calculated by the calculation unit based on the image quality of the input image and the reference image; and a fingerprint that is included in the input image based on the correlation value corrected by the correction unit. Determining means for determining whether or not the user is the user who has registered the reference image; 3. The correction means generates correction data corresponding to image qualities of the input image and the reference image from the clarity of streamline directions of the input image and the reference image, and uses the correction data to generate the correlation data. 3. The fingerprint matching device according to claim 1, wherein the value is corrected. 4. A fingerprint matching method for comparing an input image obtained by optically reading a fingerprint with a reference image registered in advance to perform fingerprint matching, wherein a correlation value between the input image and the reference image is calculated. The corrected correlation value is corrected using the clarity of the streamline direction indicating the image quality of the input image and the reference image, and based on the corrected correlation value, a user having a fingerprint included in the input image is referred to by the user. A fingerprint matching method characterized in that it is determined whether or not the user has registered the image. 5. A recording medium for use in a fingerprint matching device that compares an input image obtained by optically reading a fingerprint with a reference image registered in advance and performs fingerprint matching, wherein a correlation between the input image and the reference image is provided. Calculate the value, correct the calculated correlation value using the clarity of the streamline direction indicating the image quality of the input image and the reference image, and, based on the corrected correlation value, determine the fingerprint included in the input image. A recording medium for use in a fingerprint matching device, wherein a program for determining whether or not a user having the reference image is the user who has registered the reference image is stored.