JP2006318309A - Information processing method, information processor, computer program, and storage medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide fingerprint matching technology eliminating the necessity of advanced resources and having a high matching rate. <P>SOLUTION: An information processing method is provided with: an acquisition process for acquiring an image to be matched and user's identification information; a conversion process for applying prescribed conversion processing to the image to be matched which is acquired by the acquisition process; a reading process for reading out coordinate information corresponding to the identification information acquired by the acquisition process and matching information from a storage means in which the user's identification information, the coordinate information indicating a two-dimensional information position to be compared and the matching information of the two-dimensional information position indicated by the coordinate information are stored in relation with each other; and a judgment process for judging whether the image to be matched is an image corresponding to the identification information or not on the basis of the information of the two-dimensional information position indicated by the coordinate information of the image converted by the conversion process and the matching information corresponding to the two-dimensional information position. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a technique for identifying an individual by fingerprint verification.

  2. Description of the Related Art Conventionally, fingerprint authentication technology is known as one of so-called biometric authentication for identifying an individual using biometric information. Fingerprint authentication includes verification for identifying an individual and identification for identifying an attribute of the individual.

  Fingerprint authentication is realized by acquiring a fingerprint image using a predetermined fingerprint image acquisition sensor, performing image processing on the acquired image information, and performing identification, authentication, and the like.

  Conventionally, methods related to image processing of fingerprint images are known. For example, a pattern matching method is known as a method for directly comparing a registered fingerprint image with an input comparison fingerprint image and determining the identity thereof. According to this method, a good collation rate can be obtained in a period in which the change in the state of the finger from when the fingerprint is registered is small.

  There is also known a method of extracting feature amounts from fingerprint images and comparing the extracted feature amounts with each other instead of directly comparing the fingerprint images.

  As such a technique, a minutiae method (feature point comparison method) is known in which feature points of fingerprints (such as ridge end points and branch points are called minutiae) are compared as feature amounts. This makes it possible to obtain a highly accurate collation rate.

  Further, a frequency analysis method is known in which a section obtained by slicing a fingerprint pattern is regarded as a waveform, and a waveform spectrum series is registered and verified as a feature amount.

  As a method included in the frequency analysis method, one-dimensional group delay spectrum conversion is performed on fingerprint images with the characteristic of separating and emphasizing individual peaks of the frequency spectrum, and the obtained group delay spectrum (GDS) is used as a feature quantity. A method used for matching fingerprints is known (Patent Document 1).

In addition, as a method included in the frequency analysis method, the center position of the verification fingerprint image is detected, and the light / dark signals of a plurality of predetermined linear pixels passing through the center position are detected and converted to the time / frequency domain. A method for performing the processing is also known (Patent Document 2).
Japanese Patent No. 3057590 JP 2000-348178 A

  However, since the pattern matching method directly compares fingerprint images, the pattern matching method cannot cope with changes in the fingerprint over time, and a good matching rate cannot be obtained by a simple pattern matching method. The configuration based on the pattern matching method requires a large amount of data to be stored and acquired and requires a large capacity memory.

  The minutiae method can obtain a high accuracy collation rate, but has a drawback that it is weak against fingerprint fading and residual fingerprints because comparison is made with respect to local portions of fingerprint information. In addition, in order to perform advanced calculation processing, the configuration based on the minutiae method requires high calculation resources (processing capacity, memory amount, system cost, power consumption, etc.).

  Also, the frequency analysis method has a problem that the collation rate is not as high as that of the minutiae method.

  As described above, depending on the conventional method, a high computational resource is required to obtain a high verification rate.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a fingerprint collation technique with a high collation rate that does not require advanced resources.

According to the present invention,
An acquisition step of acquiring an image to be verified and identification information of the user;
A conversion step of performing a predetermined conversion process on the image to be collated acquired in the acquisition step;
Acquired in the acquisition step from the storage means in which the identification information of the user, the coordinate information indicating the two-dimensional information position to be compared, and the collation information of the two-dimensional information position indicated by the coordinate information are stored in association with each other. A step of reading out the coordinate information and the verification information corresponding to the identification information;
Based on the two-dimensional information position information indicated by the coordinate information of the image converted in the conversion step and the matching information corresponding to the two-dimensional information position, the image to be collated is used as the identification information. An information processing method comprising: a determination step of determining whether or not the image is a corresponding image.

  According to the present invention, it is possible to provide a fingerprint matching technique with a high matching rate that does not require advanced resources.

  Embodiments according to the present invention will be described below in detail with reference to the accompanying drawings. However, the constituent elements described in this embodiment are merely examples, and are not intended to limit the scope of the present invention only to them.

<< First Embodiment >>
(Equipment configuration)
A configuration of an apparatus that performs fingerprint image matching processing in the present embodiment will be described with reference to FIG. FIG. 3 is a diagram showing a hardware configuration of an information processing apparatus (hereinafter referred to as a fingerprint collation apparatus) that performs fingerprint image collation processing in the present embodiment.

  In FIG. 3, reference numeral 31 denotes an image sensor that captures a fingerprint image and acquires the shape of the fingerprint ridge. The image sensor 31 is realized by a system such as an optical system, a capacitance detection system, an electric field detection system, a charge injection system, and a thermal system. The imaging sensor 31 takes, for example, a two-dimensional area shape or a one-dimensional line shape. In the present embodiment, a case where a two-dimensional area shape is adopted will be described. In addition, hereinafter, for simplification of description, it is assumed that the pixel pitch of the sensor 31 is 90 μm in length and width and the number of pixels is 256 × 256 pixels = 65536 pixels, but the embodiment according to the present invention is not limited thereto. It is not done.

  Reference numeral 32 denotes a sensor control circuit that controls the operation of the image sensor 31 and operates based on the control of a CPU 36 to be described later. Reference numeral 33 denotes a frame memory for storing a frame image output from the image sensor 31.

  A data bus 34 controls the data flow in the fingerprint collation device. Reference numeral 35 denotes a bus controller that controls the operation of the data bus 34, and operates based on the control of the CPU 36 described later.

  A CPU 36 controls the operation of the entire apparatus, and executes a control program and the like stored in a memory 37 described later. For example, the CPU 36 performs processing for appropriately processing and transforming the fingerprint image information stored in the frame memory 33 and storing it in the memory 37 described later.

  Reference numeral 37 denotes a memory, which functions as a ROM for storing programs such as basic I / O programs and various data, and various data such as feature values of fingerprint images, and serves as a work area based on the control of the CPU 36. It has a function as an operating RAM. In the present embodiment, a configuration in which the memory 37 and the frame memory 33 are different will be described. However, the memory 37 and the frame memory 33 may be realized as the same memory device. The frame memory 33 and the memory 37 are generally realized by a semiconductor storage device, but may be realized by an external storage device such as a hard disk device, for example. Or media such as a flexible disk (FD), CD-ROM, CD-R, CD-RW, PC card, DVD, IC memory card, MO, memory stick, and external storage drive for realizing access to the media It may be realized by.

(Feature registration process)
The fingerprint collation apparatus according to the present embodiment registers in advance the features of a fingerprint image of a user who performs fingerprint identification or authentication (hereinafter collectively referred to as collation), and performs the fingerprint collation when performing fingerprint collation. The fingerprint collation process is executed by capturing an image and comparing the feature quantity of the obtained fingerprint image with the registered feature quantity. Therefore, next, processing for registering the feature amount of the fingerprint image will be described with reference to FIGS.

  FIG. 10 is a flowchart showing an outline of the feature amount registration processing. For convenience of explanation, the following processing is executed by an information processing device (hereinafter referred to as a registration device) such as a personal computer (PC) or a workstation (WS). May be. Similar to the fingerprint collation device, this registration device is configured as illustrated in FIG. 3 and includes an image sensor and the like.

  First, in step S1001, the registration apparatus acquires a predetermined number of fingerprint images from all users who perform fingerprint verification, respectively, via an imaging sensor.

  Next, in step S1002, for each acquired fingerprint image, a process for cutting out the fingerprint central region to be compared is performed.

  First, the entire acquired image is binarized to obtain a fingerprint ridge shape. Next, a thinning process for transforming the ridge into one line is performed. FIG. 11 is a diagram exemplarily showing a thinned ridge. In FIG. 11, 111 is a thin ridge.

Next, one or more ridges are selected from the thinned ridges. For example, a predetermined ridge is selected at random. In FIG. 11, 112 is the selected ridge.
Then, one or more points on the selected ridge are selected. For example, the selection is made at regular intervals from the selected ridge. In FIG. 11, P ₁ ,..., P ₆ are examples of selected points.

Next, for each selected point, a straight line that passes through the point and is perpendicular to the tangent line of the ridge at that point (hereinafter referred to as a perpendicular line of the ridge at the point) is obtained. For example, the tangent of the ridge in P ₁ is l _1, a straight line perpendicular to the street straight l ₁ to P _1, i.e., perpendicular ridges in P ₁ is m _1. Similarly, l ₂ and m ₂ are the tangent and perpendicular to the ridge at P _2, respectively.

Next, all intersections of the ridge normals at each selected point are determined. For example, in FIG. 11, Q ₁ is an intersection of perpendicular lines m ₁ and m ₂ . Next, the center of gravity of all intersections of perpendiculars is obtained.

  The above processing is executed a predetermined number of times, and the average value of the calculated center of gravity coordinates is set as the fingerprint center coordinate. Then, a fingerprint center region (for example, a size of 128 × 128 pixels) is cut out from the fingerprint image acquired by the image sensor 31 around the fingerprint center coordinates. Further, control is performed so that the clipped image is stored in the memory. Note that the process of cutting out the fingerprint center region is not limited to the above, and may be performed by a known method. Further, rotation correction or the like may be appropriately performed using a known method or the like.

  Next, in step S1003, all the images cut out in step S1002 are frequency-transformed once, and then subjected to inverse transformation to return to spatial coordinates to obtain an image composed of a plurality of partial images (sub-images). Alternatively, processing for obtaining an image by performing spatial resolution conversion is performed. As an example of the frequency conversion, there is a wavelet conversion used in JPEG2000. A wavelet is a local wave and is converted based on a generating function as illustrated in FIG. FIG. 5 is a diagram exemplarily showing a generating function in the wavelet transform. An example of an image format using spatial resolution conversion is a FlashPix format. In this embodiment, the case of performing primitive wavelet transform (that is, resolution conversion) is described as an example of such conversion, but the conversion method is not limited to this.

The registration device performs primitive wavelet transform on a 128 × 128 pixel 8-bit image according to the following z-transform formula.
Low (z) = (1 + z) / 2 (1)
High (z) = (1-z) / 2 (2)
The order of conversion is the same as in FIG. 4 or JPEG2000. First, the image 41 having a size of 128 squares is first applied in the direction of each horizontal line (each row) by applying the formulas (1) and (2). To separate.

  Next, the equations (1) and (2) are applied again in the direction of each vertical line (each column) in the same manner, and the low frequency low frequency component LL, the low frequency high frequency component LH, the high frequency low frequency component HL, and the high frequency high frequency component HH are applied. To separate. These four components correspond to the primary component 42 of the image 41.

  Next, the same conversion is performed only on the pixel value in the region of the low frequency component LL of the primary component 42. As a result, four secondary components 43 of the image 41 are obtained. If the same transformation is applied to the LL of the secondary component 43, the tertiary component 44 of the image 41 is obtained this time.

  By repeating the same operation seven times, the pre-conversion image 41 having a size of 128 is finally converted into a conversion image (float, 32 bits) having four seventh-order components (size 1 × 1 pixel). Is converted.

Next, in step S1004, for each user, a process of obtaining a difference amount for each pixel of the converted image obtained from the fingerprint image of the same user is performed. That is, the number of fingerprint images to be captured for the same user is M, and the converted images obtained by performing the processing of step S1003 on the fingerprint image of a certain user are I ₁ , I _{2 ,.} And At this time,
0 <i <j ≦ M (3)
For all combinations of integers (i, j) that satisfy the above, the square amount of the difference between the corresponding pixel components of I _i and I _j is obtained for all the pixel components. That is, for each converted pixel value 128 × 128 = 16384, the following amount (square difference)
(C _ik -C _jk ) ² (4)
Calculate Here, C _ik is the k th pixel value of the converted image I _i , and C _jk is the k th pixel value of the converted image I _j . The pixel value is a parameter related to the color of the pixel. For example, the pixel value is a value such as brightness, saturation, and hue, but is not limited thereto. Assuming that the number of users is N, such a set of 128 × 128 = 16384 difference squared quantities is N × _M C ₂ = N × M × (M−1) / 2. 5)
There will be only one.

Next, in step S1005, the distribution of the difference square amount calculated from the converted image of the same user is evaluated for each pixel, and the amount corresponding to the average value X _s of the difference square amount distribution and the standard deviation σ _s is determined as the pixel. Ask every time. Therefore, the amount corresponding to the average value and the standard deviation is the number of each component, that is, 128 × 128 = 16384 sets for each user.

  FIG. 8 is a diagram schematically showing the distribution of the difference squared amount. In FIG. 8, the horizontal axis represents the difference square amount value as the feature comparison amount, and the vertical axis represents the frequency of samples having the same difference square amount value. In FIG. 8, reference numeral 81 denotes a distribution of the difference square amount for a certain pixel calculated from the converted image of the same user. Reference numeral 82 denotes the distribution of the difference square amount calculated from the converted images of different users, which is obtained in steps S1006 and 1007, as will be described later.

Next, in step S <b> 1006, for each user, a process for obtaining a difference amount between the pixel of the converted image of the user and the pixel of the converted image of a different user is performed. That is, if the number of fingerprint images to be captured for the same user is M and the number of users is N, M converted images of one user and (N-1) of other (N-1) users. The square amount of the difference between the corresponding pixel components from the M converted images is obtained for all components. Such a set of 128 × 128 = 16384 difference square quantities is a total of
_N C ₂ × M × M = (N × (N−1) / 2) × M ² = M ² × N × (N−1) / 2 (6)
There will be only one.

Next, in step S1007, pixels an amount of the distribution of the differential 2 Noryou for each user calculated in step S1006 and rated for each pixel, corresponding to the average value X _d and the standard deviation sigma _d of the difference 2 Noryou distribution Ask every time. Therefore, the amount corresponding to the average value and the standard deviation is the number of each component, that is, 128 × 128 = 16384 sets for each user. Reference numeral 82 in FIG. 8 exemplarily shows the distribution of the square difference between the converted image of the user and the converted image of another user for the same pixel as the distribution 81.

  In addition, although the structure which employ | adopted the difference square amount as feature comparison amount has been demonstrated, as long as it is a value depending on the difference of a pixel value, you may comprise using another value.

Next, in steps S1008 to S1009, for each user, a process of extracting pixel components used for fingerprint identification and authentication from 16384 sets of (X _s , σ _s , X _d , σ _d ) is performed.

here,
1) The distance between the comparison distribution 81 between the same fingerprints and the comparison distribution 82 between the non-identical fingerprints is separated,
2) It can be said that the pixel component that satisfies the characteristic that the overlap between both distributions is small has a low probability of confusion between the person and the other person, and is considered to be a good fingerprint feature amount. Therefore, in step S1008, a determination value D as shown below is calculated for 16384 sets of (X _s , σ _s , X _d , σ _d ).
D = | X _d −X _s | / (3 × (σ _s + σ _d )) (7)
The numerator is an amount corresponding to the distance between both distributions described in 1). The denominator represents the breadth of both distributions. When the widths of both distributions are widened, the two distributions overlap, and the collation rate, that is, the person rejection rate, the other person acceptance rate, and the like deteriorate.

Next, in step S1009, a predetermined number K of pixel components are selected in descending order of the determination value D, and the coordinates (x ₁ , y ₁ ), (x ₂ , y ₂ ) of the selected pixel components,. ·, and stores and holds as (x _K, y _K) the feature vector coordinates (position coordinates). Alternatively, the coordinates of a pixel component having a D value equal to or greater than a certain value may be selected as the feature vector coordinates.

  FIG. 6 is a diagram exemplarily showing the determination value D in the region where 0 ≦ x ≦ 15 and 0 ≦ y ≦ 15. In FIG. 6, the horizontal direction is the x direction, the vertical direction is the y direction, and the upper left is the origin (0, 0). In FIG. 6, the coordinates of a pixel component having a determination value D of 0.03 or more are selected as feature vector coordinates. In FIG. 6, a term indicated by a shaded portion is a pixel component that indicates a value of D of 0.03 or more. There are a total of 69 in FIG. In this example, it is assumed that there are coordinates where the value of 1 D is 0.03 or more in an unillustrated area, and there are a total of 70 (69 + 1) in the entire 128-square image. In this example, pixel components corresponding to K = 70 coordinates extracted in this way are used for comparison as fingerprint feature values.

In this embodiment, the fingerprint collation is determined by comparing the value obtained by weighting each component of the feature vector with the following formula and the predetermined threshold.
Ctot = Σαk × (Crk−Csk) ² (8)
Here, Csk is the average value of the pixel values of the converted image at the coordinates corresponding to the k-th component of the feature vector coordinates extracted in step S1009. Crk is a pixel value at a corresponding coordinate extracted from a fingerprint image to be compared. The (linear) coefficient αk is a coefficient applied to the difference square term. Σ is taken for a total of K feature vector components k. In this case, if the coefficient αk is set larger as the component k better represents the feature of the fingerprint, the reliability of feature quantity comparison can be improved.

  Therefore, in step S1010, processing for optimizing the coefficient αk of the feature vector extracted in step S1009 is performed. Any method may be adopted as a method for optimizing the coefficient αk, but here, two methods will be described as examples.

  One is a method using the judgment value (7) of the equation (7). It can be said that the larger the determination value D shown in equation (7), the better the component represents the characteristics of the fingerprint. Therefore, a large value is assigned to the coefficient of such a component.

For example, if the value of D is written as D (x, y), it is clearly shown that it is a function of 128 × 128 components, and the linear coefficient αk is expressed as αk (x, y) = D (x, y).・ (9)
And That is, it is considered that the larger the value of D (x, y), the more the characteristic of the fingerprint appears. Therefore, this is reflected in the calculation of the fingerprint comparison amount Ctot expressed by the equation (8). is there.

  The other method is to measure a change in collation rate when the coefficient αk is changed in advance using a fingerprint library, and obtain an optimum combination of linear coefficients αk by the least square method or the like from the result.

For example, if the value of Ctot (equation (8)) in the same fingerprint comparison distribution is Cs and the value of Ctot in the non-identical fingerprint comparison distribution is Cd,
Cd (ak) / Cs (ak) = max (Cd / Cs) (10)
Should hold. Here, ak is an ideal value of αk.
Differentiating this by each ak,
δ (Cd / Cs) / δak = ((C′d × Cs−Cd × C ′s) / Cs ² ) = 0 (11)
∴ C′d × Cs−Cd × C ′s = 0 (12)
Should hold.

  The value of αk can be optimized by obtaining αk using the equation (12). These values only need to be optimized once during the registration process, and are not processes that particularly require processing time or newly require resources during fingerprint matching. The storage of the linear coefficient αk only requires at most 70 × 4 = 280 bytes.

  For example, the collation rate can be increased by optimizing αk by the above method.

  Next, in step S1011, the value obtained by the equation (8) is compared with the magnitude to obtain a threshold value used for fingerprint collation determination. The derivation of the threshold will be described with reference to FIG. FIG. 9 is a diagram exemplarily showing a distribution of fingerprint feature amounts (difference squared amounts) when fingerprint comparison is performed.

  In FIG. 9, the horizontal axis, the vertical axis, and 81 and 82 are the same as those in FIG. In FIG. 9, Vth is a threshold value used for determination of fingerprint collation for one of the K pixels selected in step S1009. Reference numeral 91 denotes an area of the same fingerprint comparison distribution 81 in which the difference square amount is larger than the threshold value Vth. In the area 91, the principal is rejected, that is, the fingerprint of the user whose legitimate user is not legitimate. It will be determined. On the other hand, reference numeral 92 denotes an area of the non-identical fingerprint comparison distribution 82 in which the difference square amount is smaller than the threshold value Vth. In the area 92, another person is accepted, that is, a user whose legitimate user's fingerprint is valid. It will be determined as a fingerprint. Therefore, it is desirable to select the threshold value Vth so that the areas of the region 91 and the region 92 are reduced. Therefore, for example, the difference square amount at the point where the comparison distribution 81 between the same fingerprints and the comparison distribution 82 between the non-identical fingerprints intersect is set as the threshold value Vth.

Threshold values Vth (V1, V2,..., VK) are obtained for all the K pixels selected in step S1009 by the method exemplified above. Then, a value Vtot obtained by the following expression is set as a threshold value of the sum of squared differences Ctot.
Vtot = Σαk × (Crk−Csk) ² (13)
However, the meanings of Crk, Csk, and αk are the same as described above. It should be noted that the value of Vtot may be set to a predetermined value according to the application and purpose after repeating the experiment.

  After the above processing, the registration device sends information necessary for the fingerprint verification device to perform fingerprint verification (hereinafter referred to as verification information) to the fingerprint verification device and stores it in the memory 37 of the fingerprint verification device. To do. The collation information includes the registered user's user ID, feature vector coordinates, threshold value Vtot, count αk (α1,..., ΑK), and average value Csk (Cs1,..., CsK) of pixel values of the converted image. Included and related to each other. It should be noted that the collation information to be transmitted only needs to include information on users who perform fingerprint collation using the fingerprint collation apparatus, and it is not necessary to transmit collation information for all users. In this case, the collation information calculated in the registration device is moved to the memory 37 of the fingerprint collation device by, for example, a predetermined communication means.

  The collation information does not necessarily need to be stored and held in the memory 37 of the fingerprint collation device. For example, the collation information may be stored and held in an external device and used by the fingerprint collation device with reference to it when necessary. For example, a configuration in which a storage device that can communicate with the fingerprint verification device stores and holds verification information, and the fingerprint verification device and the storage device communicate to perform fingerprint verification during fingerprint verification. Alternatively, the fingerprint collation device and the registration device may be connected via a network, and the fingerprint collation device and the registration device may communicate to perform fingerprint collation during fingerprint collation.

  When the fingerprint collation device communicates with an external device, the fingerprint collation device includes a communication interface that can exchange data with the external device. However, the communication interface here is not only an interface with a data transmission / reception line such as LAN, wireless LAN, Internet, public line (analog line, ISDN, etc.), WAN, etc., but also a flexible disk (FD), You may implement | achieve as an external storage drive for implement | achieving access to media, such as CD-ROM, CD-R, CD-RW, PC card, DVD, IC memory card, MO, and a memory stick.

(Fingerprint verification process)
Next, a fingerprint collation process in which the fingerprint collation apparatus reads fingerprint information and collates the fingerprint information will be described. FIG. 1 is a flowchart showing the flow of fingerprint collation processing.

  First, in step S <b> 101, the fingerprint collation apparatus acquires a fingerprint image from a user who collates fingerprints via an imaging sensor. The acquired fingerprint image is stored in the frame memory 33 as a comparison image. Here, it is assumed that the comparison image is acquired with a gradation value of 256 (8 bits).

  Next, in step S102, a fingerprint center region to be compared is extracted from the comparison image acquired in step S101. This process is the same as the process in step S1002 of the feature amount registration process.

  Next, in step S103 and step S104, the frequency of the image cut out in step S102 is once converted, and then inverse conversion is performed to return to spatial coordinates to obtain an image composed of a plurality of partial images (sub-images). I do. This process is the same as the process in step S1003 of the feature amount registration process.

  Next, in step S105, the pixel value at the position coordinates (xi, yi) stored in the memory 37 as feature vector coordinates is extracted from the converted comparison image. A combination (vector expression) of the extracted pixel values becomes a feature vector of the comparison image. In the present embodiment, as described above, the total number of pixel values is K = 70. Therefore, the feature vector consists of 70 components (32 bits each).

  Next, in step S106, it is determined whether or not the comparison with the feature vector of the comparison image has been performed for all the feature vectors stored in the memory 37 of the fingerprint collation apparatus. If so (YES in step S106), the process proceeds to step S110. If not (NO in step S106), the process proceeds to step S107. When evaluating step S106 for the first time, if data of one or more feature vectors is stored in the memory 37, the evaluation result in step S106 is NO, and the process proceeds to step S107.

  In step S107, one of the feature vectors stored in the memory 37 is selected, and the calculation of equation (8) is performed on this feature vector and the feature vector of the comparison image to obtain the value of Ctot. That is, Csk is a value at the kth coordinate of the feature vector selected from the memory 37, Crk is a value at the kth coordinate of the feature vector of the comparison image, and αk is a coefficient corresponding to the selected feature vector. Ask. Here, it is assumed that a feature vector having a user ID i is selected.

  Next, in step S108, the Ctot value obtained in step S107 is compared with the threshold value Vtot associated with the feature vector selected in step S107. If Ctot ≦ Vtot holds, it is determined that the characteristics of the registered fingerprint image match the characteristics of the input image (YES in step S108), and the process proceeds to step S109. If Ctot> Vtot, it is determined that the registered fingerprint image feature and the input image feature do not match (NO in step S108), the process returns to step S106, and the process is continued.

  In step S109, it is determined that the comparison image acquired by the imaging sensor is the same as the fingerprint of the registered image, and another system in which the fingerprint verification system dominates that the registered fingerprint having the user ID i matches the comparison fingerprint. Is notified using the CPU 36. If necessary, this result is also notified to the system user and displayed. Then, the process ends.

  When the process returns to step S106, it is determined whether or not the comparison with the comparison image has been performed for all the data stored in the memory 37. If all the steps have been performed (YES in step S106), the process proceeds to step S110. In step S110, a predetermined system is notified that fingerprint collation is impossible (failure), and if necessary, the user is informed that collation is impossible, and a series of processing is terminated.

  If the comparison with the comparison image has not been performed for all the data stored in the memory 37 (NO in step S106), the process proceeds to step S107. In step S107, one of the feature vectors stored in the memory 37 that has not yet been selected is selected (for example, the user ID is j), and the feature vector stored in the memory 37 is selected. One of these is selected, and the calculation of the equation (8) is performed on the feature vector and the feature vector of the comparison image to obtain the value of Ctot. The processing after step S108 is as described above.

  The processes in steps S103 and S104 can be executed by using a known conversion algorithm that does not require advanced resources. In addition, the process of step S107 only performs the squaring, multiplication, and addition at most K (= 70) times. Furthermore, the process of step S108 is a simple comparison of numbers. Further, the number of loops in steps S106 to S108 is only the maximum number of feature vectors stored in the fingerprint collation apparatus. Therefore, the above process can be executed without requiring high computational resources. In the case of using a file structure in which the sub-image and the fingerprint feature amount are already embedded in the file format like the FlashPix file format, most of the processes in steps S103 and S104 are not required. .

  In this embodiment, since K = 70 and the data amount of one pixel is 32 bits, the size of the feature vector (registered image) for one person registered in the fingerprint collation apparatus is 70 × 32 = 2240. Bit = 280 bytes. As described above, since the size of the feature vector per person is very small, the above configuration can be applied to a device system without a large-capacity memory.

  In addition, by repeatedly performing frequency conversion and inverse conversion in steps S103 and S104, variations in image brightness and contrast can be prevented, and accurate image matching can be performed. Further, as described in (feature amount registration processing), by selecting a feature vector coordinate having a high determination value D and setting a coefficient αk based on the determination value D, the fingerprint image is configured. Can be compared with particular emphasis on the characteristic parts of the, and the matching rate can be increased.

<< Second Embodiment >>
In the first embodiment, the configuration using the primitive wavelet transform exemplified by the equations (1) and (2) has been described, but any transformation equation may be used. For example, a wavelet transform used in JPEG2000 as shown in equations (14) to (17) may be used.
R0 (z) = H0 (z) × X (z) (14)
R1 (z) = H1 (z) × X (z) (15)
H0 (z) = (− z ⁻² + 2 × z ⁻¹ + 6 + 2 × z−z ² ) / 8 (16)
H1 (z) = (− z ⁻¹ + 2−z) / 2 (17)
Here, X (z) is an input signal. R0 (z) is a low frequency output (low frequency output), R1 (z) is a high frequency output (high frequency output), and H0 (z) and H1 (z) are integer type 5/3 filters. Inverse transformation is also easily obtained, but is omitted here.

  Hereinafter, processing of the registration device and the fingerprint verification device corresponding to the present embodiment will be described with reference to FIGS. The configuration of the present embodiment is substantially the same as the configuration of the first embodiment, and here, differences from the configuration of the first embodiment will be described.

(Feature registration process)
First, processing in steps S1001 to S1002 is performed. These contents are almost the same as the processing in the first embodiment. However, DC level shift processing is performed when the fingerprint central region to be compared is cut out from the captured image. The DC level shift process is a process of subtracting a positive value from the tone value of all the pixels so that the center of gravity of the tone value of the image becomes zero. Since the size of the fingerprint image to be acquired is assumed to be as small as 128 × 128 pixels, the tile division used in JPEG 2000 is not particularly required. In the present embodiment, each pixel is represented by an 8-bit integer type.

In step S1003, instead of the conversions of the expressions (1) and (2), the conversions of the expressions (14) to (17) are performed. Then, a thinning process is executed. The signal after thinning is expressed by the following equations (18) and (19).
S0 (z) = {H0 (z ^1/2 ) X (z ^1/2 ) + H0 (−z ^−1/2 ) X (−z ^−1/2 )} / 2 (18)
S1 (z) = {H1 (z ^1/2 ) X (z ^1/2 ) + H1 (−z ^−1/2 ) X (−z ^−1/2 )} / 2 (19)
The fractions that appear here are rounded, but their reversibility is guaranteed. These processes are performed up to the 7th order in the same manner as in the first embodiment. Next, inverse conversion is performed on each converted image to return pixel values from frequency coordinates to spatial coordinates. The processing is interpolation and inverse wavelet transform. As a result, the data structure illustrated in FIG. 4 is obtained. However, the data amount of each pixel value slightly increases from 8 bits (256 gradations) to 9 bits due to the previous DC level shift.

  The processes after step S1004 are the same as those in the first embodiment.

(Fingerprint verification process)
First, the processes of steps S101 and S102 are performed. These contents are almost the same as the processing in the first embodiment. However, as in the feature amount registration process, the DC level shift process is performed when the fingerprint center region to be compared is cut out from the captured image. Also in this process, the tile division used in JPEG 2000 is not particularly required.

  In step S103 and step S104, similarly to the first embodiment, the image cut out in step S102 is once subjected to frequency conversion, and then subjected to inverse conversion to return to spatial coordinates, from a plurality of partial images (sub-images). A process for obtaining an image is performed. This process is the same as the process in step S1003 of the feature amount registration process.

  The processes after step S105 are the same as those in the first embodiment.

  In this embodiment, since the component constituting the feature vector is an integer type, the information amount of the registered feature vector is 70 × 9 bits = 630 bits = 79 bytes compared to the first embodiment adopting the float type. It can be considerably reduced.

  As described above, the wavelet transform formula is not limited to those exemplified in the formulas (1) and (2). Of course, the present invention is not limited to those exemplified in the equations (14) to (19). It is known that there are various wavelet transforms in the world, and can be freely selected according to the use and purpose. For example, it may be configured by using various wavelet transform software such as “Wavelet Explorer” (product name).

  Further, the frequency conversion is not limited to the wavelet transform, and if the frequency conversion has a thinning operation for frequency conversion and an interpolation operation for inverse conversion as described above, the same functions and effects as in the present embodiment can be obtained. it can.

<< Third Embodiment >>
In the embodiment according to the present invention, as the comparison image and the registered image, a file format in which a plurality of images having different degrees of resolution already converted, such as the above-described FlashPix, may be used. In this case, without performing the conversion operation in steps S103, S104, S1003, etc., the first, the above-described first, Fingerprint images can be compared as in the second embodiment. Further, the method using resolution conversion is not limited to the method using FlashPix, and any method can be adopted depending on the application and purpose.

  In this embodiment, since the image before conversion is stored and used as a registered image in the memory 37, the information amount of the registered image (including the feature vector and an extra part) is slightly increased, but registration (fingerprint) is used. It is also possible to display the image itself on a predetermined display device.

  When resolution conversion is used instead of frequency conversion, the fingerprint collation apparatus performs processing in accordance with a flowchart as shown in FIG. 2 in which steps S103 and S104 in FIG. 1 are replaced with S203. Note that the registration apparatus performs resolution conversion processing in step S1003 of FIG.

<< Fourth Embodiment >>
In step S1002 and step S102 of the first embodiment, the processing for cutting out the fingerprint central region to be compared from the acquired fingerprint image is performed. However, before or after the processing for cutting out, the acquired fingerprint image is processed. The image processing may be performed as appropriate.

  For example, the acquired comparative fingerprint image may be expanded and contracted in a predetermined direction. This example will be described with reference to FIG. FIG. 7 is a diagram illustrating an example in which the comparative fingerprint image 71 acquired by the imaging sensor 31 is expanded and contracted in the longitudinal direction. Reference numeral 72 denotes a fingerprint image expanded by + 5% in the longitudinal direction. In the case of using the stretched image, for example, a total of five images expanded by -2%, -1%, 0%, + 1%, + 2% from the acquired comparative image are created, and the feature vector obtained from them is Compare with registered feature vector. Thereby, even when the skin of a finger having a fingerprint expands or contracts for some reason, it is possible to correctly collate. Of course, it may be performed not only in the longitudinal direction but also in the lateral direction. In addition, the above-described operation may be performed for the first time when the collation cannot be performed without performing the expansion / contraction operation first.

In addition, as other preprocessing and image processing, the following image processing may be performed.
1) Rotation correction 2) Binarization 3) Luminance adjustment 4) Contrast and gain adjustment 3) includes the DC level shift described in the second embodiment. By using the above image processing together, the collation rate can be further increased.

<< Other Embodiments >>
As described above, the exemplary embodiments of the present invention have been described in detail. However, the present invention can take an embodiment as, for example, a system, an apparatus, a method, a program, or a storage medium. You may apply to the system comprised from an apparatus, and may apply to the apparatus which consists of one apparatus.

  The present invention can also be achieved by supplying a program that realizes the functions of the above-described embodiment directly or remotely to a system or apparatus, and the computer of the system or apparatus reads and executes the supplied program code. Including the case where it is achieved.

  Therefore, since the functions of the present invention are implemented by a computer, the program code installed in the computer is also included in the technical scope of the present invention. That is, the present invention includes a computer program itself for realizing the functional processing of the present invention.

  In that case, as long as it has the function of a program, it may be in the form of object code, a program executed by an interpreter, script data supplied to the OS, or the like.

  As a recording medium for supplying the program, for example, floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card ROM, DVD (DVD-ROM, DVD-R) and the like.

  As another program supply method, a client computer browser is used to connect to an Internet homepage, and the computer program of the present invention itself or a compressed file including an automatic installation function is downloaded from the homepage to a recording medium such as a hard disk. Can also be supplied. It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention.

  In addition, the program of the present invention is encrypted, stored in a storage medium such as a CD-ROM, distributed to users, and key information for decryption is downloaded from a homepage via the Internet to users who have cleared predetermined conditions. It is also possible to execute the encrypted program by using the key information and install the program on a computer. In addition to the functions of the above-described embodiments being realized by the computer executing the read program, the OS running on the computer based on an instruction of the program is a part of the actual processing. Alternatively, the functions of the above-described embodiment can be realized by performing all of them and performing the processing.

  Furthermore, after the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion board or The CPU or the like provided in the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

It is a flowchart which shows the flow of a fingerprint collation process. It is a flowchart which shows the flow of the fingerprint collation process in 3rd Embodiment. It is the figure which showed the hardware constitutions of the fingerprint collation apparatus in this embodiment. It is the figure which showed typically the order of the wavelet transformation in 1st Embodiment. It is the figure which showed the generating function in a wavelet transform exemplarily. It is the figure which showed determination value D value about the area | region of 0 <= x <= 15, 0 <= y <= 15. It is the figure which showed the example which expands and contracts the comparison fingerprint image acquired by the imaging sensor. It is the figure which showed typically distribution of the difference square amount. It is the figure which showed exemplarily the distribution of the difference squared amount at the time of performing fingerprint comparison. It is a flowchart which shows the outline of the registration process of the feature-value. It is the figure which showed the ridgeline thinned thinly.

Claims (12)

An acquisition step of acquiring an image to be verified and identification information of the user;
A conversion step of performing a predetermined conversion process on the image to be collated acquired in the acquisition step;
Acquired in the acquisition step from the storage means in which the identification information of the user, the coordinate information indicating the two-dimensional information position to be compared, and the collation information of the two-dimensional information position indicated by the coordinate information are stored in association with each other. A step of reading out the coordinate information and the verification information corresponding to the identification information;
Based on the two-dimensional information position information indicated by the coordinate information of the image converted in the conversion step and the matching information corresponding to the two-dimensional information position, the image to be collated is used as the identification information. A determination step of determining whether or not the image is a corresponding image.   The information processing method according to claim 1, wherein the predetermined conversion process is at least one of frequency conversion, frequency reverse conversion, resolution conversion, and resolution reverse conversion.   The information processing method according to claim 1, wherein the predetermined conversion process is a wavelet transform.   The information processing method according to any one of claims 1 to 3, wherein the conversion step performs the predetermined conversion process a plurality of times. The storage means further stores determination information and weighting information of the two-dimensional information position indicated by the coordinate information in association with the identification information of the user,
The determination step is based on information on the two-dimensional information position indicated by the coordinate information of the image converted in the conversion step, and the matching information and the weighting information corresponding to the two-dimensional information position. The comparison information is derived, and based on a comparison between the comparison information and the determination information, it is determined whether or not the image to be collated is an image corresponding to the identification information. 5. The information processing method according to any one of 4 above.   The comparison information is obtained by squaring a difference value between the information of the two-dimensional information position indicated by the coordinate information of the image converted in the conversion step and the matching information corresponding to the two-dimensional information position. 6. The information processing method according to claim 1, wherein the obtained value is a sum of values obtained by multiplying the obtained value by the weighting information corresponding to the two-dimensional information position. Further comprising a generating step of generating the coordinate information;
The generating step includes
For each of a plurality of images acquired from a first user and a plurality of images acquired from one or more other users, the conversion process is performed, respectively.
Obtaining a statistic of each two-dimensional information position of the plurality of images of the first user subjected to the conversion process as a first statistic;
Obtaining a statistic of each two-dimensional information position of the plurality of images of the first and other users subjected to the conversion process as a second statistic;
7. The two-dimensional information position suitable for collation is selected as the coordinate information based on the relationship between the first statistic and the second statistic. 7. Information processing method. The generating step includes
The information processing method according to claim 7, wherein the weighting information is obtained based on the first statistic and the second statistic. An image processing step of performing image processing including at least one of expansion / contraction operation, rotation correction, binarization, luminance adjustment, contrast adjustment, and gain adjustment on the image acquired in the acquisition step;
The information processing method according to claim 1, wherein the conversion step performs a predetermined conversion process on the image that has been subjected to the image processing in the image processing step. Acquisition means for acquiring an image to be verified and identification information of the user;
Conversion means for performing a predetermined conversion process on the image to be collated acquired by the acquisition means;
Storage means for associating and storing user identification information, coordinate information indicating the two-dimensional information position to be compared, and collation information of the two-dimensional information position indicated by the coordinate information;
Reading means for reading out the coordinate information and the collation information corresponding to the identification information acquired in the acquisition means from the storage means;
Based on the two-dimensional information position information indicated by the coordinate information of the image converted by the conversion means and the matching information corresponding to the two-dimensional information position, the image to be collated is used as the identification information. An information processing apparatus comprising: a determination unit that determines whether the image is a corresponding image.   A computer program for causing a computer to execute the information processing method according to claim 1.   A computer-readable storage medium storing the computer program according to claim 11.

Images